The study, published in Communications Engineering, was led by Jinchao Liang and Ze Zhang of the Aerospace Information Research Institute at the Chinese Academy of Sciences, together with colleagues from the University of Chinese Academy of Sciences, Aerospace Information University, and partner institutions. Their method, which they call pixel-segmenting imaging, rests on a deceptively simple observation: a pixel does not respond uniformly to light across its area. Quantum efficiency, the probability that an incoming photon generates an electron that the readout circuit can count, varies from point to point within each pixel because of the way microlenses, color filters, and silicon doping are laid down during fabrication. If those variations can be mapped precisely, each pixel becomes a tiny array of detectors with different sensitivities, and the raw sensor output becomes a mixture that can be mathematically unscrambled.

Unscrambling that mixture requires knowing the sensitivity map with high accuracy, and this is where the team’s custom optical system and evolutionary algorithms come in. The researchers illuminate the sensor with carefully designed optical patterns and measure how each pixel responds, then use an evolutionary optimization procedure to infer the intra-pixel quantum efficiency distribution. Evolutionary algorithms, which mimic natural selection by breeding candidate solutions and keeping the fittest, are well suited to this kind of inverse problem because the search space of possible sensitivity maps is enormous, nonlinear, and resistant to conventional gradient-based fitting. By iterating through generations of candidate maps and refining them against measured responses, the algorithm converges on a physical characterization of the sensor that would be extremely difficult to obtain by direct measurement alone.

Once the intra-pixel sensitivity map is known, the imaging model changes fundamentally. The light falling on a single pixel is no longer summed into one value; instead, the measured output is treated as a weighted combination of the intensities reaching each of the sixteen sub-regions, with the weights given by the measured quantum efficiencies. With a custom optical system that projects appropriately structured illumination onto the sensor, the team can solve for those sub-region intensities, effectively reconstructing the scene at a resolution sixteen times finer in area than the native pixel grid. The minimal sampling unit drops from one pixel to one-sixteenth of a pixel, which means the sensor can recover high-frequency spatial information that the original images, under conventional sampling, could not contain.

The performance numbers reported in the paper are striking. At a cut-off frequency of 25 line pairs per millimetre, a standard benchmark for imaging sharpness, the modulation transfer function of the reconstructed images reaches 2.3 times that of the original images. The structural similarity index measure, a metric that captures how faithfully an image preserves perceptually meaningful structure, improves by a factor of 3.1 relative to the originals. These are not marginal gains; they represent a substantial recovery of information that conventional sampling theory would suggest is unrecoverable from a given sensor.

What makes the result particularly notable from an engineering standpoint is that it requires no modification of the sensor hardware itself. The pixel-segmenting approach works with the original device structure intact, relying on a custom optical system in front of the sensor and computational reconstruction behind it. This stands in contrast to strategies that shrink pixels, stack them in three dimensions, or add micromirror arrays, all of which demand expensive changes to fabrication processes. Because the method characterizes the sensor as it is, it could in principle be applied to existing cameras, turning a software-and-optics upgrade into a resolution upgrade.

The distinction between this approach and deep-learning super-resolution is worth emphasizing. Neural-network upscalers are trained on large datasets and learn statistical priors about what natural images look like; when they add detail, they are generating plausible content rather than measuring it. The pixel-segmenting method, by contrast, derives its extra resolution from a physical property of the sensor, the spatial variation of quantum efficiency within each pixel, that is measured and verified for the specific device in use. The recovered high-frequency information is therefore grounded in actual photon measurements rather than in learned expectations, which matters for scientific and security applications where fabricated detail is worse than no detail at all.

The implications reach across several domains. In remote sensing, where satellite and aerial cameras are weight- and power-constrained and cannot simply carry larger sensors, squeezing more resolution out of existing pixels could sharpen Earth-observation imagery without new hardware. In microscopy and industrial inspection, where the finest resolvable feature often determines what can be detected, a sixteen-fold increase in effective sampling density could reveal structures previously blurred away. Consumer imaging could benefit as well, since smartphone cameras are already pushing against the physical limits of pixel miniaturization; a computational layer that extracts sub-pixel information would let manufacturers improve sharpness without the penalty of smaller, noisier pixels.

There are, of course, practical considerations that will shape adoption. The method depends on an accurate intra-pixel quantum efficiency map, which must be measured for each sensor or sensor type, and the evolutionary optimization procedure requires computational resources and careful optical calibration. The custom optical system that structures the illumination must also be integrated into the imaging path, which may be straightforward for dedicated instruments but more challenging for mass-market devices. How the approach performs under real-world conditions, with varying illumination, motion, and noise, will determine whether the laboratory gains translate into field performance. The published work demonstrates the principle and quantifies the improvement under controlled conditions, providing the foundation on which such engineering questions can now be addressed.

For a field that has largely assumed the pixel is an indivisible unit of measurement, the study reframes the problem in a way that is likely to attract attention well beyond optical engineering. By showing that the sampling constraint is not the pixel boundary but the uniformity assumption within it, the researchers have opened a path that combines precision optical characterization, evolutionary computation, and image reconstruction into a single pipeline. If the approach proves robust across sensor types and imaging conditions, the humble pixel, the workhorse of digital imaging for half a century, may turn out to have been hiding sixteen detectors in plain sight all along.

Subject of Research: Sub-pixel imaging using intra-pixel quantum efficiency measurement to exceed pixel-level sampling limits in digital image sensors

Article Title: Pixel-segmenting imaging physically exceeds pixel-level sampling constraints via intra-pixel quantum-efficiency measurement enabled by evolutionary algorithms

Article References: Liang, J., Xue, H., Zhang, C., Guan, M., Xiao, H., Sun, C., Liu, Y., Liu, X., Gao, L., & Zhang, Z. (2026). Pixel-segmenting imaging physically exceeds pixel-level sampling constraints via intra-pixel quantum-efficiency measurement enabled by evolutionary algorithms. Communications Engineering. https://doi.org/10.1038/s44172-026-00773-7

Image Credits: AI Generated

DOI: 10.1038/s44172-026-00773-7

Keywords: digital image sensors, super-resolution, quantum efficiency, evolutionary algorithms, pixel-segmenting imaging, modulation transfer function, optical engineering, computational imaging, sampling constraints, high-frequency recovery, Communications Engineering, image reconstruction

The research addresses a stubborn gap in construction management science. Traditional time–cost trade-off models, which have dominated scheduling research since the era of the Critical Path Method, focus almost exclusively on minimizing duration and expenditure. Quality, despite being a decisive factor in structural performance, durability, safety and client satisfaction, has usually been reduced to a subjective rating or evaluated only after the schedule is fixed. The problem is that the three objectives are genuinely in conflict: compressing a schedule demands overtime, extra crews and parallel workstreams that inflate costs, while cutting corners on materials or labour to save money erodes the very quality that makes a building last. Solving this three-way tug-of-war mathematically requires an algorithm that can explore an enormous space of possible schedules without collapsing prematurely onto a mediocre answer.

The researchers, led by GVLN Murthy of Swarnandhra College of Engineering and Technology, attacked the problem in two stages. First, they needed a way to make quality measurable. They turned to the Analytic Hierarchy Process, or AHP, a well-established decision-making technique that converts qualitative expert opinions into precise numerical weights through structured pairwise comparisons. A panel of five professionals — two academics, two senior project managers and a quality consultant, each with 17 to 24 years of experience — compared the importance of thirteen construction activities and three quality indicators: labour performance, material quality and equipment efficiency. The judgments were aggregated with the geometric mean and checked for logical consistency using Saaty’s consistency ratio, which all comparisons passed. The result was a Project Quality Index, a single number that captures how well a given schedule performs on quality, weighted by how much each activity actually matters to the building’s integrity.

Notably, the AHP analysis revealed that foundation work dominated the weighting: footing received the highest priority at 0.190, followed by excavation at 0.164 and groundworks at 0.160, while finishing activities scored far lower because they contribute less to structural stability. This weighting scheme means the optimizer automatically learns that skimping on the foundation is far more damaging to overall quality than economizing on paint and trim — exactly the kind of engineering common sense that has been so hard to encode in scheduling software before.

The second stage is where the artificial intelligence comes in. The team employed the Non-dominated Sorting Genetic Algorithm III, a state-of-the-art evolutionary optimizer designed for problems with multiple conflicting objectives. NSGA-III works by evolving a population of candidate schedules, using a reference-point mechanism to keep the solutions spread evenly across the trade-off surface rather than clustering in one corner. But conventional NSGA-III has a known weakness: on complex combinatorial problems it can converge too early, settling into a local optimum and missing better solutions elsewhere in the search space. To fix this, the researchers added Opposition-Based Learning, a technique that evaluates not only each candidate solution but also its mirror image across the search space. By keeping whichever of the two performs better, the algorithm dramatically enriches the initial population and accelerates its march toward the best regions of the solution landscape.

The team tuned the algorithm carefully through a grid search combined with sensitivity analysis. A population of 100 individuals evolving over 200 generations, with a crossover probability of 0.90, a mutation probability of 0.10 and an opposition-learning rate of 50 percent, produced the best balance between exploration and computational cost. Simulated binary crossover and polynomial mutation were selected after outperforming their alternatives. The framework was then validated on a genuine project: a ground-plus-one reinforced concrete residential building in Gwalior, Madhya Pradesh, with a built-up area of roughly 420 square metres, thirteen activities linked by finish-to-start precedence relationships, and two to four execution modes per activity, each with its own duration, cost, resource demands and quality score.

The results were striking. The optimizer generated 23 Pareto-optimal scheduling alternatives — schedules in which no objective can be improved without worsening another. Project durations ranged from 165 to 200 days, total costs from about USD 493,529 to USD 765,747, and quality index values from roughly 0.85 to 0.91. The fastest schedule, at 165 days, came at a steep price of USD 737,439, while the cheapest option saved money but stretched the timeline to 186 days with lower quality. The highest-quality schedule pushed costs to their maximum. Crucially, the algorithm consistently assigned premium execution modes to structurally critical activities such as basement and exterior wall construction, while choosing economical options for non-critical finishing work — a resource allocation pattern that mirrors what an experienced engineer would do, but derived automatically from the mathematics.

When the researchers applied the Weighted Sum Method with stakeholder weights of 0.4 for time, 0.4 for cost and 0.2 for quality, Solution 14 emerged as the preferred compromise: 181 days, USD 528,572 and a quality index of about 0.901. Sensitivity analysis confirmed the choice was robust — shifting the weights toward time or cost favoured different solutions, but the rankings changed only marginally when the AHP quality weights were varied by plus or minus 10 percent, and the algorithm’s performance held steady across variations in population size and mutation settings.

The head-to-head comparison is perhaps the most convincing evidence. Benchmarked against NSGA-III, Multi-Objective Particle Swarm Optimization, Multi-Objective Ant Colony Optimization, Opposition-Based Multi-Objective Differential Evolution and a Latin-Hypercube-Sampling-enhanced NSGA-III over 30 independent runs each, the proposed framework achieved the best scores on nearly every metric: the most unique non-dominated solutions (23), the highest hypervolume (0.92), the lowest generational distance (1.39) and the lowest error ratio. Wilcoxon signed-rank and Friedman tests confirmed the improvements were statistically significant at the 0.05 level. The only trade-off was computational time, which at 81 seconds ran slightly longer than some competitors — a modest price for a demonstrably richer and better-converged set of options.

The authors are candid about the limitations. The validation rested on a single residential project with deterministic durations and costs, and the AHP weights, however carefully checked, ultimately rest on human judgment. But the framework’s modular design leaves the door open to stochastic scheduling, fuzzy AHP, Monte Carlo simulation, resource-constrained variants, Building Information Modelling, digital twins, sustainability objectives and reinforcement learning. For an industry where schedule overruns and quality failures cost billions annually, the message is clear: the era of treating quality as an unquantified footnote in construction planning may finally be ending, replaced by optimization engines that weigh every trade-off transparently and let decision-makers see exactly what each day, dollar and quality point is worth.

Subject of Research: Multi-objective optimization of time, cost and quality trade-offs in construction project scheduling using AHP and opposition-based NSGA-III

Article Title: An intelligent multi-objective decision framework for Time–Cost–Quality optimization in construction scheduling using AHP and OBL–NSGA-III

Article References: Murthy, G., Prasad, R., Venkatakrishna, A., Sahoo, S. K., Mahalik, S., Sahoo, B. P., Dash, B., & Potharaju, S. (2026). An intelligent multi-objective decision framework for Time–Cost–Quality optimization in construction scheduling using AHP and OBL–NSGA-III. Discover Artificial Intelligence, 6(1), Article 1470. https://doi.org/10.1007/s44163-026-02456-w

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02456-w

Keywords: construction scheduling, multi-objective optimization, NSGA-III, opposition-based learning, Analytic Hierarchy Process, time-cost-quality trade-off, Pareto-optimal solutions, evolutionary algorithms, project management, Project Quality Index, decision support, artificial intelligence

The new method, called MOPSO-NMF, is the work of Mehdi Rahimi, Keyhan Khamforoosh, and Vafa Maihami, and it fuses two well-established but previously separate strands of computational intelligence. The first is multi-objective particle swarm optimization, a swarm-based search technique inspired by the way flocks of birds collectively explore a landscape. The second is non-negative matrix factorization, or NMF, a linear algebra tool that decomposes large matrices of non-negative numbers into lower-rank components. NMF has a long history in machine learning, famously used to learn the parts of objects in images and topics in document collections, and it has been applied to community detection before. What makes the new work distinctive is the specific role it assigns to NMF: rather than using it as the community detector itself, the researchers use it to construct the starting population for the evolutionary search.

To understand why that matters, it helps to look at how standard evolutionary community detection algorithms operate. These methods maintain a population of candidate solutions, each one a proposed partition of the network’s nodes into communities. The population evolves over generations as better solutions are selected, recombined, and mutated, guided by one or more quality functions such as modularity. The trouble, the authors argue, lies at the very beginning of this process. Most evolutionary algorithms initialize their populations randomly, which means the first generation of candidate partitions is essentially noise. On a noisy, explicit network topology, such random starts frequently lead to premature convergence, where the search locks onto a mediocre solution early and never escapes, and to poorly defined community boundaries where ambiguous nodes are assigned arbitrarily.

MOPSO-NMF sidesteps this problem by never searching the raw topological space at all. Instead, the algorithm first applies non-negative matrix factorization to the network’s adjacency representation, extracting a low-rank latent representation of the network. In this compressed latent space, each node is described by its membership strengths across a set of latent factors, and the underlying block structure of the network, obscured by noise in the original topology, becomes far more visible. The researchers then use this latent representation to generate structure-aware initial particles for the particle swarm. In other words, the swarm does not begin its journey from random guesses; it begins from educated hypotheses about where the communities are, hypotheses derived directly from the statistical structure of the network itself.

The multi-objective side of the framework matters just as much. Community detection is inherently a trade-off problem: one quality function may reward densely connected groups while another rewards well-separated ones, and optimizing a single measure can distort the final partition. Multi-objective evolutionary algorithms handle this by searching for a whole family of compromise solutions, the Pareto-optimal set, from which a decision maker can later select. By running the particle swarm in this multi-objective mode, MOPSO-NMF explores the trade-off surface between competing objectives rather than committing to a single, possibly misleading, definition of what a good community looks like. The swarm’s particles move through the latent space, guided by personal and collective experience, gradually assembling a front of high-quality candidate partitions.

One further innovation refines the details that matter most in practice: the boundary nodes. In almost every real network, the vast majority of nodes are easy to classify because they sit deep inside a community, surrounded by their own kind. The difficult cases are the nodes at community borders, which maintain ties to several groups at once and are the usual source of errors in any partition. MOPSO-NMF incorporates a boundary-based local search operator that specifically targets these ambiguous nodes. Working in the latent space, the operator re-examines the assignments of border nodes and adjusts them when a better placement is available, polishing the partitions produced by the swarm without disturbing the parts of the solution that are already well settled.

The experimental evaluation is where the approach shows its teeth. The researchers tested MOPSO-NMF on a battery of synthetic and real-world benchmark networks, comparing it against fifteen state-of-the-art community detection algorithms. On the Karate club network, the famous social network of a university karate team that split into two factions, and on the Dolphin network of bottlenose dolphin associations in Doubtful Sound, New Zealand, the algorithm achieved a normalized mutual information score of exactly 1, meaning it recovered the known community structure perfectly. On the Football network of American college football games, it scored 0.9547, and on the Political Books network of political titles linked by co-purchase patterns, it reached 0.8430. Precision, recall, and F-measure scores told the same story: perfect 1.00 across all three measures on Karate and Dolphin, 0.95, 0.97, and 0.96 on Football, and 0.94, 0.95, and 0.95 on Political Books.

Those numbers are more than a leaderboard. Normalized mutual information is a stringent measure that quantifies how much information the detected partition shares with the ground-truth partition, and a score of 1 indicates exact agreement. The fact that a single framework can hit that ceiling on multiple networks while remaining competitive on harder, noisier benchmarks suggests that the three ingredients, NMF-based initialization, multi-objective optimization, and boundary refinement, are genuinely complementary rather than redundant. The initialization gives the swarm a head start informed by the network’s latent structure, the multi-objective search keeps the exploration balanced across competing quality criteria, and the local search cleans up exactly the nodes where automated methods typically stumble.

The broader significance lies in a lesson that resonates across computational intelligence: where a search starts often matters as much as how it searches. Population initialization has long been recognized as a critical factor in the performance of metaheuristics, and a growing body of literature shows that informed, deterministic, or hybrid initialization strategies can dramatically accelerate convergence and improve solution quality. By importing a matrix factorization technique from the machine learning toolbox into the initialization stage of a swarm algorithm, the Iranian team illustrates a productive convergence between the two fields. NMF provides the statistical intuition; the evolutionary framework provides the global search power and the flexibility to juggle multiple objectives; and the boundary operator supplies the fine-grained local polish that global methods lack.

There are, of course, practical questions that remain open. The authors note that the code used to implement the algorithm is available from the corresponding author upon reasonable request, and the benchmark datasets are publicly available, which should make independent replication straightforward. Scaling structure-aware initialization to networks with millions of nodes, handling overlapping communities where nodes genuinely belong to several groups at once, and adapting the latent representation to dynamic networks whose structure changes over time are all natural next steps. But the core result stands on its own: when the search for hidden communities begins from a representation that already understands the network, rather than from blind random guessing, the communities come into focus with remarkable clarity. For a problem that underpins everything from recommendation systems to epidemiology, that is a starting point worth paying attention to.

Subject of Research: Community detection in complex networks using a multi-objective particle swarm optimization algorithm initialized with non-negative matrix factorization

Article Title: A multi-objective evolutionary algorithm based on non-negative matrix factorization for community detection in complex networks

Article References: Rahimi, M., Khamforoosh, K., & Maihami, V. (2026). A multi-objective evolutionary algorithm based on non-negative matrix factorization for community detection in complex networks. Neural Computing and Applications, 38(19), Article 804. https://doi.org/10.1007/s00521-026-12599-7

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12599-7

Keywords: community detection, complex networks, non-negative matrix factorization, multi-objective optimization, particle swarm optimization, evolutionary algorithms, Pareto front, modularity, benchmark networks, machine learning, graph clustering, swarm intelligence

The core insight behind the work is that classification errors across classes can conflict. A network that excels at distinguishing shirts from coats in the Fashion MNIST dataset might struggle with sneakers versus sandals, and the conventional practice of averaging errors into a single scalar loss hides these tensions entirely. By contrast, the CMaO-SL formulation defines one objective per class, so a ten-class problem becomes a ten-objective optimization problem. Because no single architecture can simultaneously minimize every class-specific error, the search targets a Pareto set: a collection of non-dominated solutions in which improving performance on one class necessarily comes at some cost to another. This trade-off information, normally invisible to practitioners, becomes an explicit output of the training process.

Solving such a problem is computationally daunting. Not only must the search evaluate many candidate architectures, but it must do so repeatedly to draw statistically meaningful conclusions about a stochastic process. The authors tackled this challenge with a two-pronged strategy combining transfer learning with evolutionary neural architecture search. Every network in their framework is split into two parts: a deep, pre-trained feature extractor kept completely frozen, and a shallower classification sub-network that is allowed to evolve. The feature extractor is derived from a ResNet-18 network pre-trained on ImageNet, with its final layers removed and its pooling adjusted to output 2,304 features. Because the backbone is never retrained or fine-tuned, the expensive part of the computation is paid only once, and the search can concentrate on the smaller, cheaper classification head.

The evolutionary search itself adapts MOEA/D, a well-established decomposition-based multiobjective evolutionary algorithm, to the architecture search setting. The user first defines a master architecture for the classification sub-network, consisting of fully connected layers with batch normalization and ReLU activations. Each individual in the evolving population is a binary mask vector that dictates which connections in the master architecture are pruned; where a mask entry is zero, the corresponding connection is removed and its weight fixed at zero. This pruning-as-search strategy generates a rich family of alternative architectures from a single blueprint. Each candidate is then trained with gradient descent, evaluated per class on a validation set, and assigned a performance vector whose components are one minus the F1 score for each class.

The evaluation procedure embeds an early-stopping mechanism to guard against overfitting. During each candidate’s five training epochs, the algorithm tracks a scalar validation error and retains the best-performing weight vector, ensuring that later epochs cannot degrade the network that gets assessed. The authors used two training loss functions in their experiments, cross-entropy and a weighted average of per-class binary cross-entropy aligned with each sub-problem’s reference direction, and two corresponding validation error functions. This produced five distinct search configurations, three evolutionary and two non-evolutionary, the latter serving as controls in which independent scalar searches replaced the coupled evolutionary process.

The experimental study spanned six benchmark datasets drawn from very different domains: Fashion MNIST, K-MNIST, CIFAR10, and three medical image collections from the MedMNIST suite, namely BloodMNIST, DermaMNIST and OCTMNIST. Each configuration was run fifteen times, allowing a rigorous statistical treatment that is rare in the Pareto-based architecture search literature. After confirming with a Shapiro-Wilks test that the results were not normally distributed, the authors compared configurations using medians and pairwise Mann-Whitney-Wilcoxon tests at a significance level of 0.05, alongside the IGD+ performance indicator, which measures how well an obtained front approximates a reference Pareto front.

The results were consistent across datasets and evaluation sets. Networks trained with cross-entropy loss produced better-approximated Pareto fronts than those trained with the weighted binary cross-entropy variant, and the evolutionary searches matched or beat their non-evolutionary counterparts on front quality while producing sets of networks with lower total complexity. Parallel coordinate plots of per-class F1 scores made the conflicting nature of the objectives vividly visible: the line connecting a classifier’s best performance on one class typically dipped on others. Crucially, when the authors picked the single best network from each evolutionary run’s set, measured by overall validation accuracy, those networks matched the accuracy of the unpruned master architecture while being substantially smaller, demonstrating that the many-objective formulation finds leaner models without sacrificing performance.

The benchmark comparisons are perhaps the most striking part of the study. When the networks discovered by the proposed method were plotted against published reference results for ResNet-18-based classifiers on the same datasets, no published result dominated the new networks on the combined axes of accuracy and complexity. The full ResNet-18 carries roughly 11.5 million parameters relevant to the comparison, whereas the evolved networks came in at around 2.5 million, meaning the reference architecture is about 360 percent more complex. That dramatic reduction in size cost less than 8 percent in accuracy, and on the DermaMNIST skin-lesion dataset the evolved classifiers actually surpassed the published benchmark. For a method that never fine-tuned its feature extractor and trained each candidate for only five epochs, this is a remarkable showing.

The practical implications extend beyond raw benchmark numbers. Because the output of the search is a set of classifiers with exposed trade-offs, users gain flexibility that single-model pipelines cannot offer. A deployment with tight computational budgets can select the smallest acceptable network from the set, while an application where a particular class matters most, say a rare disease category in a medical screening tool, can pick the member that favors that class. Alternatively, the trade-off structure can guide the construction of ensembles, an avenue the authors identify as a priority for future work. They also flag several open questions: whether fine-tuning the shared feature extractor would push accuracy higher, how the approach handles imbalanced data and noisy labels common in real-world problems, and how alternative pruning schemes, such as removing whole layers or blocks, might perform.

The study also fills a conceptual gap in the literature. Although Pareto-based multiobjective machine learning has existed for years, formal treatments of many-objective supervised learning, where the number of objectives exceeds three, have been scarce, and no prior work had combined such a formulation with neural architecture search and a transferred feature extractor. Most existing multiobjective NAS studies juggle only two or three objectives, typically accuracy, model size and inference cost, rather than one objective per class. By demonstrating a working pipeline at ten objectives, backed by fifteen-run statistical analysis per configuration, the Tel-Aviv and Braude team has shown that class-aware, many-objective architecture search is not only feasible but competitive. As automated machine learning matures, approaches like this suggest a future in which the question is no longer simply how accurate a network is, but which compromises it makes, and whether those are the compromises its users actually want.

Subject of Research: Class-based many-objective neural architecture search for deep classifiers using evolutionary optimization and transfer learning

Article Title: Class-based many-objective neural architecture search for deep classifiers

Article References: Mardiks, G., Salih, A., & Moshaiov, A. (2026). Class-based many-objective neural architecture search for deep classifiers. Neural Computing and Applications, 38(19), Article 800. https://doi.org/10.1007/s00521-026-12510-4

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12510-4

Keywords: neural architecture search, many-objective optimization, Pareto front, evolutionary algorithms, transfer learning, deep learning, multi-class classification, MOEA/D, pruning, ResNet-18, AutoML, machine learning

The core difficulty that CMOEA-TSD targets is deceptively simple to state: in many constrained problems, the set of solutions that satisfy all constraints — the feasible region — is tiny, fragmented, or shaped like a narrow winding corridor through a vast infeasible expanse. Classical multi-objective evolutionary algorithms, which mimic natural selection by evolving a population of candidate solutions toward a set of optimal trade-offs known as the Pareto front, tend to stumble in such terrain. Populations converge prematurely onto whatever small pocket of decent solutions they first encounter, losing the genetic diversity needed to explore elsewhere. Worse, when the feasible region is sparse, a single population must simultaneously explore the broader landscape, satisfy constraints, and maintain a well-spread approximation of the Pareto front — three demands that pull the search in incompatible directions and often lead to stagnation.

The new algorithm, developed by Weiye Bai, Cunyan Liu, and Wei Wang, attacks this problem by refusing to make one population do everything. Instead, CMOEA-TSD runs three cooperating subpopulations in parallel, each with a distinct mandate. The first is a global search component that approximates the unconstrained Pareto front — the optimal trade-off surface one would obtain if constraints were ignored entirely. This may seem like a detour, but the unconstrained front provides a valuable map of where the best objective values lie, even if those points violate constraints. The second subpopulation performs a diversity-enhanced search: it reformulates the original problem into a relaxed multi-objective form using reference vectors and a dynamically shrinking constraint threshold, effectively treating constraint violation as an additional objective that is gradually tightened over time. The third subpopulation is feasibility-oriented, focusing squarely on driving solutions into the constraint-satisfying region.

This division of labor is orchestrated through a two-stage structure. In the early stage of the search, the algorithm emphasizes exploration: the diversity-enhanced and unconstrained-front populations roam widely, gathering information about the landscape’s structure and locating promising regions that a purely feasibility-driven search would never visit. As the run progresses, a stage-switching mechanism — driven by the overall population’s convergence rate — adaptively adjusts mating strategies, shifting the emphasis toward exploitation. In practical terms, the algorithm first learns where the interesting parts of the search space are, then concentrates its reproductive effort on refining solutions within and near the feasible region. The convergence-rate trigger is crucial: rather than switching stages at a fixed generation count, the algorithm monitors how quickly the population is actually improving, allowing the transition to happen at the right moment for each individual problem.

The technical machinery behind the diversity-enhanced subpopulation deserves particular attention. Reference vectors, a concept popularized by decomposition-based evolutionary algorithms, partition the objective space into directional cones, ensuring that the search maintains coverage across the entire Pareto front rather than clustering on a few segments. By pairing these vectors with a constraint threshold that shrinks dynamically, the algorithm implements a gradual tightening strategy: solutions with moderate constraint violations are tolerated early on because they may carry useful genetic material that points toward the feasible region, but the tolerance is progressively reduced until only fully feasible, well-distributed solutions remain. This approach echoes a growing body of evidence in the evolutionary computation literature that infeasible solutions are not merely waste — they can be essential stepping stones, particularly when the feasible ratio of a problem is vanishingly small.

To validate the design, the authors subjected CMOEA-TSD to an unusually thorough experimental regimen. The algorithm was tested on four benchmark test suites — standardized collections of constrained multi-objective problems engineered to expose specific weaknesses, such as large infeasible regions, deceptive landscapes, and low feasible ratios. It was then applied to nine real-world engineering problems, the acid test for any optimization method, since industrial problems bring irregular constraint structures and noisy objective landscapes that synthetic benchmarks often fail to capture. Across these evaluations, CMOEA-TSD was compared against seven state-of-the-art algorithms, and the results showed significant performance advantages for the new method, both in converging toward the true Pareto front and in maintaining a well-spread set of solutions along it.

Just as important as the headline results are the ablation studies, in which the authors systematically removed components of their algorithm to verify that each part earns its place. These experiments confirmed that both the three-population co-evolutionary structure and the diversity enhancement mechanisms contribute measurably to performance. This kind of component-level validation matters in a field where new algorithms sometimes owe their wins to a handful of clever tricks whose individual contributions are unclear. By demonstrating that the multi-population architecture and the relaxed-search strategy each pull their weight, the study offers a clearer picture of why the method works, not just that it works — a distinction that matters for researchers hoping to build on the approach.

The significance of this work extends beyond the evolutionary computation community. Constrained multi-objective optimization sits at the heart of countless practical applications: flow-shop scheduling for energy savings, industrial copper burdening, optimal power flow in electrical grids, and cleaner production planning have all been framed in these terms in recent literature. When an algorithm fails to find feasible, well-distributed trade-off solutions, engineers are left either violating constraints or accepting inferior designs. A method that reliably handles sparse feasible regions could therefore translate into tangible gains — lower energy consumption, better resource allocation, and more robust engineering designs — across sectors where multiple objectives and hard constraints collide.

The study also reflects a broader trend in the field: the move from single-population, single-phase algorithms toward co-evolutionary frameworks in which specialized populations exchange information and adapt their roles over time. Earlier work explored two-population and dual-archive designs, push-and-pull search strategies, epsilon-constraint handling, and helper-problem-assisted multitasking. CMOEA-TSD pushes this line of thinking further with its tri-population structure and convergence-based stage switching, suggesting that the future of constrained optimization may lie less in any single clever operator and more in architectures that coordinate multiple search behaviors intelligently. The authors, who received no external funding for the study, report no conflicts of interest, and the algorithm’s components — reference vectors, shrinking thresholds, and adaptive mating — are all implementable within standard evolutionary computation platforms.

For practitioners, the message is that the hardest constrained problems — those with narrow, disconnected feasible regions and strongly conflicting objectives — are becoming tractable with the right architectural choices. For researchers, the paper offers a template: divide the search burden, protect diversity as a first-class citizen rather than an afterthought, and let measured convergence rather than arbitrary schedules dictate when to shift from exploring to exploiting. As computational optimization continues to permeate engineering, logistics, and machine learning system design, algorithms like CMOEA-TSD illustrate how a deeper understanding of search dynamics — when to spread out, when to converge, and how to let specialized populations cooperate — can turn previously intractable problems into solvable ones. The full study is available in Cluster Computing, and its benchmark code and experimental protocols provide a foundation for the next generation of constrained multi-objective solvers.

Subject of Research: A multi-population evolutionary algorithm for constrained multi-objective optimization

Article Title: A novel multi-population evolutionary algorithm based on two-stage division and diversity enhancement for constrained multi-objective optimization

Article References: Bai, W., Liu, C., & Wang, W. (2026). A novel multi-population evolutionary algorithm based on two-stage division and diversity enhancement for constrained multi-objective optimization. Cluster Computing, 29(12), Article 732. https://doi.org/10.1007/s10586-026-06515-w

Image Credits: AI Generated

DOI: 10.1007/s10586-026-06515-w

Keywords: evolutionary algorithms, constrained multi-objective optimization, Pareto front, diversity enhancement, co-evolution, constraint handling, reference vectors, stage switching, benchmark test suites, engineering optimization, Cluster Computing, Xi'an Polytechnic University

The problem the researchers set out to solve sits at the intersection of two challenging branches of evolutionary computation. Multi-objective optimization asks an algorithm to find not one best answer but an entire frontier of trade-off answers — the Pareto front — when goals conflict, as they almost always do in engineering design. Multi-task optimization goes further, demanding that several related problems be solved simultaneously within a single run, with useful knowledge flowing between them so that progress on one task accelerates progress on the others. Combine the two and you get multi-objective multi-task optimization, or MO-MTO, a field that has grown rapidly because so many real-world scenarios — scheduling a factory while tuning its production line, designing a component while optimizing its manufacturing process — naturally present themselves as families of coupled problems.

Most existing MO-MTO algorithms, however, share two structural blind spots. First, they typically rely on a single genetic operator to generate offspring, meaning the crossover and mutation rules applied to parent solutions never change as the population evolves. This is a serious handicap, because the search landscape of an optimization problem changes character dramatically over time: aggressive, exploration-heavy operators are valuable early on, when the population is scattered and the algorithm needs to cover ground, while fine-grained, exploitation-focused operators matter later, when the population has converged near the Pareto front and precision matters more than reach. A fixed operator is a compromise that is rarely optimal at any stage. Second, when these algorithms transfer knowledge between tasks, they confine that transfer to the objective space — the space where solutions are scored — and ignore the decision space, the space of the actual variables being optimized, where much of the most valuable structural information about a problem resides.

MTEA-RLKS attacks both weaknesses at once. Its authors, led by Xuanwei Zhang, Zhiyuan Pan, and Long Chen of Beijing University of Chemical Technology, together with Yushu Du of the University of Melbourne and Junze Zhu of Tongji University, designed a collaborative knowledge transfer mechanism that draws on both spaces simultaneously. In the objective space, the algorithm extracts what the team calls static knowledge: statistical summaries of the global population distribution and of local neighborhoods, capturing where high-quality solutions sit relative to one another. This static knowledge is stable enough to be shared across tasks without much risk of misleading the recipient. In the decision space, by contrast, the algorithm tracks dynamic knowledge — the way the population’s configuration shifts as evolution proceeds — using Gaussian processes, a class of probabilistic models that can represent distributions over functions and quantify uncertainty about how the population will move next.

The division of labor between static and dynamic knowledge is the conceptual heart of the method. Static knowledge, extracted from global distributions and local neighborhoods, provides a coarse map of where promising regions lie, which can be safely exported to a related task even when the two problems differ in detail. Dynamic knowledge, modeled with Gaussian processes, is more like a weather forecast for the evolving population: it tells the algorithm not just where the population is, but how it is trending, allowing the transferred information to be adapted to the recipient task’s current stage of convergence. By synergizing the two, the algorithm guides the generation of offspring that are more likely to be high quality, accelerating the optimization process on every task in the portfolio rather than trading one task’s performance against another’s.

The second pillar of the framework is the part most likely to resonate beyond the evolutionary computation community: a Deep Q-Network, the same architecture that made headlines when it learned to play Atari games from raw pixels, is used to select genetic operators adaptively. In reinforcement learning terms, the state is the current situation of the evolving population, the actions are the available genetic operators, and the reward reflects how much the chosen operator improves the solutions. Crucially, the network is trained online, during the iterations of the evolutionary run itself, so its policy adapts to the evolutionary needs of different solutions as they arise. Early in a run, the network can learn to favor exploratory operators; as the population converges, it can shift toward operators that refine solutions along the Pareto front. No human expert needs to specify a schedule — the algorithm discovers one.

This marriage of deep reinforcement learning with evolutionary search is more than a technical convenience. Adaptive operator selection has been studied for years, usually through bandit-style methods or hand-crafted rules that track simple statistics such as operator success rates. A Deep Q-Network brings something richer: the ability to recognize complex, high-dimensional patterns in the population state and to learn long-horizon strategies rather than myopic one-step choices. In a multi-task setting this matters enormously, because the state of one task’s population can legitimately influence what operator is best for another task, and a deep value function is far better positioned to capture such cross-task dependencies than a running average of success counts.

The empirical case for the method rests on the two standard benchmark suites for this problem class, CEC2017 and CEC2019, which contain collections of multi-objective multi-task problems designed by the evolutionary computation community specifically to stress knowledge transfer and operator choice. Across these test sets, MTEA-RLKS was compared against six state-of-the-art algorithms, and the reported results show that it significantly outperformed all of them. The benchmarks are demanding precisely because they include pairs of tasks with varying degrees of relatedness — some pairs share useful structure, others are nearly independent — so an algorithm that transfers knowledge indiscriminately can be actively harmed by its partner task. The collaborative mechanism’s blend of cautious static transfer and adaptive dynamic transfer appears to navigate that hazard well.

The practical implications stretch across the domains where multi-objective multi-task problems arise naturally. Engineering design routinely involves optimizing several coupled objectives — weight, strength, cost, energy efficiency — across a family of related variants of a product, and an algorithm that can solve the whole family at once, transferring lessons between variants, offers real savings in computation and engineering time. Scheduling, logistics, neural architecture search, and resource allocation all present similar structures. The funding acknowledgments behind the work, including China’s National Key Research and Development Program and the National Natural Science Foundation, signal that the application ambitions are substantial, spanning projects in intelligent manufacturing and information technology.

There are, of course, the usual caveats that accompany any new algorithmic result. The Deep Q-Network adds computational overhead and hyperparameters of its own, and online training during an evolutionary run raises questions about stability that the benchmarks address but real deployments will probe further. The Gaussian process models of dynamic knowledge, while elegant, carry their own scaling considerations as decision dimensions grow. Yet the direction is unmistakable: the boundary between evolutionary computation and deep reinforcement learning, long treated as separate traditions, is becoming a productive seam. MTEA-RLKS demonstrates that a neural network can serve as the decision-making brain of an evolutionary algorithm, choosing its reproductive tools moment by moment while the algorithm itself ferries knowledge between problems in two spaces at once. If the approach generalizes beyond the benchmarks as well as its authors hope, the fixed-operator evolutionary algorithm may soon look like a relic — a one-tool craftsman in a world that has learned to carry a full toolbox and, better still, to know which tool to reach for.

Subject of Research: Adaptive operator selection using deep reinforcement learning for decomposition-based multi-objective multi-task evolutionary optimization

Article Title: Adaptive operator selection via deep reinforcement learning for decomposition-based multi-objective multi-task optimization

Article References: Zhang, X., Pan, Z., Du, Y., Zhu, J., & Chen, L. (2026). Adaptive operator selection via deep reinforcement learning for decomposition-based multi-objective multi-task optimization. Complex & Intelligent Systems. https://doi.org/10.1007/s40747-026-02387-0

Image Credits: AI Generated

DOI: 10.1007/s40747-026-02387-0

Keywords: evolutionary algorithms, multi-objective multi-task optimization, reinforcement learning, Deep Q-Network, adaptive operator selection, knowledge transfer, Gaussian processes, Pareto front, CEC2017 benchmarks, CEC2019 benchmarks, decomposition-based optimization, machine learning

To appreciate why this matters, it helps to understand how conventional active noise control, or ANC, is usually designed. In the classical workflow, engineers first decide where the secondary sources, the loudspeakers that emit the anti-noise, will be placed. Only then do they estimate the secondary path, the acoustic and electronic route that a signal travels from the controller to the microphone that measures the result. That estimated path is essential, because the controller must know how its output will be transformed by the environment before it can compute the correct anti-noise signal. The problem is that in multi-speaker, multi-microphone systems, these secondary paths are strongly coupled: each speaker’s output reaches every error microphone, so identifying all the paths becomes computationally expensive, and errors in one path contaminate the estimates of the others.

Kong and Wu sidestep this bottleneck by reframing the entire problem as an optimization task. Rather than treating speaker placement as a fixed design decision made before the control system ever runs, they treat the spatial coordinates of the secondary sources and their initial phases as optimization variables. The goal of the optimization is to minimize the sound pressure levels in designated target zones, the regions of space where quiet is desired. The open-field acoustic environment is handled through boundary conditions, which allows the framework to model how sound radiates and interferes in a realistic setting without the heavy computational machinery of full path identification.

This reformulation is elegant in concept but brutal in practice. The control variables form a high-dimensional space, since every speaker contributes both position coordinates and a phase value, and the objective function that maps any candidate configuration to the resulting sound pressure in the target zones is decidedly nonconvex. In plain terms, the landscape of possible solutions is riddled with local minima, valleys that look like good answers to a naive search algorithm but are far from the true optimum. A purely gradient-based optimizer, dropped into such a landscape, will slide downhill into the nearest valley and get stuck. A purely stochastic optimizer may wander forever without ever exploiting the fine structure of the terrain near a good solution.

The researchers’ response is a hybrid solver they call the Interior Point Method-Assisted Mean Differential Evolution with Weibull distribution, abbreviated IPMDEW. The algorithm has two engines. The first is a mean differential evolution algorithm with Weibull distribution, or MDEW, which handles global exploration. Differential evolution is a population-based evolutionary technique that generates new candidate solutions by combining weighted differences between existing members of the population, allowing it to probe the search space broadly and escape the traps of local minima. The Weibull distribution, a flexible statistical distribution whose shape can be tuned, is used to guide the algorithm’s variation operators, and the mean-based formulation helps stabilize the search across the population.

The second engine is the interior point method, a classical optimization technique famous for solving constrained problems efficiently by tracing a path through the interior of the feasible region rather than bumping along its boundaries. In IPMDEW, the interior point method is not deployed from the start. Instead, it waits in the wings while the differential evolution population roams the search space, and as the population approaches convergence, the interior point method steps in to perform gradient-based refinement. This division of labor delivers what the authors describe as global-local optimization synergy: the evolutionary phase supplies the raw global insight about where the good regions of the search space lie, and the interior point phase polishes those candidates to high precision using local derivative information that the evolutionary search alone cannot exploit efficiently.

The practical payoff was demonstrated through joint simulations run in MATLAB coupled with COMSOL, a commercial multiphysics simulation platform capable of modeling acoustic wave propagation with high fidelity. This coupling matters because it means the optimization was not tested against a simplified analytical toy model but against a realistic acoustic simulation of an open-field environment, complete with the boundary conditions that govern how waves behave in such a space. In these simulations, the proposed method achieved an average noise reduction of 17.2 decibels, a figure the authors report as superior to the result obtained by the MDEW algorithm on its own. For context, a 17-decibel reduction represents a dramatic perceptual change: because perceived loudness scales roughly logarithmically with sound pressure, such a reduction can make a noisy zone sound many times quieter.

The significance of the work lies as much in its conceptual shift as in its numbers. By adopting what the authors call an optimization-driven perspective, the study treats the physical configuration of the noise control system, not just its electronic parameters, as something the algorithm itself can discover. This is a departure from the traditional pipeline in which human engineers fix the geometry first and the mathematics second. It also avoids the need to identify the secondary path explicitly, which the authors identify as a major source of computational complexity in centralized ANC systems due to the strong coupling between secondary source paths. In a centralized architecture, where one controller coordinates all the speakers, that coupling multiplies quickly as the number of sources and sensors grows, and the identification burden grows with it.

The research also sits at an interesting intersection of disciplines, something reflected in the formal classifications attached to the paper: optimal control, evolutionary algorithms, and acoustics. The mathematical machinery draws on nonlinear programming, specifically interior point methods that have transformed constrained optimization since their popularization in the 1980s, and on stochastic search techniques inspired by evolutionary biology. The application domain is classical acoustics, governed by the wave equation and the physics of interference. Making these communities talk to each other is precisely what allows the new framework to handle a problem that is simultaneously high-dimensional, nonconvex, constrained by physics, and evaluated through expensive simulations.

There are, of course, caveats that come with any simulation-stage advance. The results reported are from joint MATLAB-COMSOL simulations rather than hardware experiments in a physical room, and real-world deployment would introduce challenges that simulations approximate imperfectly, from loudspeaker nonlinearities to sensor noise to the time-varying nature of real noise fields. The paper itself is an open-access publication released on 31 August 2026, received on 25 March 2025 and accepted on 10 August 2026, and was supported by the Fundamental Research Funds for the Central Universities under Grant No. 2232026A4. The authors, affiliated with the College of Information and Intelligent Science and the Engineering Research Center of Digitized Textile and Apparel Technology at Donghua University, declare no competing financial interests. Even with those caveats, the study offers a compelling template for the next generation of noise control systems: ones in which the placement and phasing of every anti-noise speaker is not a guess made on a blueprint, but a decision computed, refined, and optimized by an algorithm that knows both how to explore the wide world of possible configurations and how to squeeze the last decibel out of the best one it finds.

Subject of Research: Optimization of secondary source placement and initial phases in centralized active noise control using a hybrid differential evolution and interior point method algorithm

Article Title: An interior point assisted differential evolution algorithm for centralized active noise control

Article References: Kong, W., & Wu, C. (2026). An interior point assisted differential evolution algorithm for centralized active noise control. Complex & Intelligent Systems. https://doi.org/10.1007/s40747-026-02472-4

Image Credits: AI Generated

DOI: 10.1007/s40747-026-02472-4

Keywords: active noise control, differential evolution, interior point method, optimal control, acoustics, sound pressure level, secondary source placement, evolutionary algorithms, COMSOL simulation, MATLAB, nonconvex optimization, Donghua University

At the heart of the problem lies a fundamental tension. An effective intrusion detection system cannot simply maximize a single measure of performance. It must simultaneously achieve high detection accuracy, minimize the rate of false alarms, maintain strong relevance between selected features and real threats, and often satisfy additional constraints such as computational efficiency and energy consumption. Improving one of these objectives frequently degrades another. A detector tuned aggressively for accuracy may flood security analysts with false positives, while a system optimized to reduce false alarms may miss subtle attacks. Traditional single-objective optimization, which collapses all these concerns into one weighted score, struggles to capture the trade-offs that real-world network defenders must navigate.

Multi-objective optimization algorithms offer a fundamentally different approach. Rather than producing a single solution, these methods search for a set of Pareto-optimal solutions, meaning configurations in which no objective can be improved without worsening at least one other. This gives security engineers a family of candidate detectors, each representing a different balance of accuracy, false alarm rate, and other criteria, from which they can select the design best suited to a particular network environment. The review highlights that swarm intelligence algorithms, inspired by the collective behavior of bird flocks, ant colonies, and other natural systems, along with evolutionary algorithms modeled on biological selection, have emerged as particularly efficient tools for exploring these complex trade-off landscapes in IoT settings.

The technical appeal of these algorithms becomes clearer when one considers the structure of the intrusion detection pipeline itself. Most modern systems operate in two stages. First, a feature selection phase identifies the most informative subset of network traffic attributes, discarding redundant or irrelevant measurements that would otherwise inflate computational cost and confuse the classifier. Second, a classification phase, increasingly built on machine learning and deep learning models, labels traffic as benign or malicious. Both stages involve search problems with competing objectives. Feature selection must balance the number of selected features against detection quality, while classifier tuning must weigh accuracy against false alarm rates and, in resource-constrained IoT devices, against processing overhead. Multi-objective optimizers can address both stages within a unified framework.

The IoT environment makes these challenges especially acute. Unlike conventional enterprise networks, IoT deployments combine devices with wildly varying computational capabilities, communication protocols, and security postures. Many smart sensors operate on batteries and cannot afford the overhead of heavyweight monitoring software. Traffic patterns are heterogeneous, and attack surfaces expand constantly as new device types join the network. The review emphasizes that intrusion detection objectives in this context must be chosen carefully among factors including false alarm rate, relevance, and detection accuracy, and that the datasets used to evaluate candidate systems play a decisive role in determining whether laboratory results translate into real-world protection.

By conducting a comparative examination of multi-objective intrusion detection approaches in the IoT, organized by their algorithmic strategies and the datasets used in their evaluation criteria, the authors provide a benchmark that both seasoned researchers and newcomers to cybersecurity can use. This kind of systematic synthesis matters because the field has grown explosively, with numerous research concepts proposed that combine swarm algorithms, evolutionary computation, deep learning, and classical machine learning techniques. Without a structured comparison, it is difficult to know which combinations of optimizer, detector architecture, and dataset genuinely advance the state of the art and which merely report favorable results on narrow benchmarks.

The review does not shy away from the field’s shortcomings. The authors identify a series of current challenges and outline future avenues for research, aiming to shed light on the gaps that remain to be filled when developing intrusion detection systems for IoT networks. Among the implicit lessons is that evaluation practices across the literature remain uneven, with studies differing in the datasets they use, the objectives they prioritize, and the metrics they report, making direct comparison difficult. The authors position their survey as a representative analysis of the period from 2018 to 2025, a window during which multi-objective methods matured from academic curiosities into practical contenders for deployment in network defense.

For the broader cybersecurity community, the significance of this work extends beyond the IoT niche. The digital revolution and the expansion of networked operations have amplified the importance of intrusion detection across all critical infrastructure, and the trade-off structures that multi-objective optimization resolves in IoT networks mirror those found in cloud data centers, 5G core networks, and industrial control systems. Techniques validated in the IoT context, where resource constraints are harshest, may prove readily transferable to environments with more generous budgets but higher stakes. The review illustrates that multi-objective optimization holds promising potential to improve several competing intrusion detection objectives simultaneously in IoT networks, a claim that, if borne out by subsequent deployment studies, could reshape how security vendors design their detection engines.

The authors, affiliated with Ajman University in the United Arab Emirates, Al-Balqa Applied University in Jordan, and Chulalongkorn University in Thailand, frame their contribution explicitly as a resource for the research community and novice cybersecurity researchers seeking to understand and develop effective intrusion detection models. Published open access on 27 September 2026, the article arrives at a moment when the volume of data transferred between connected communication devices continues to grow massively, and when the consequences of undetected intrusions, from ransomware in hospitals to sabotage of industrial systems, have never been higher. By mapping the objectives, strategies, and datasets that define the multi-objective approach to intrusion detection, the review provides a common vocabulary for a field that will need every tool it can muster as the Internet of Things continues its relentless expansion.

Subject of Research: Multi-objective optimization algorithms for intrusion detection systems in Internet of Things networks

Article Title: Multi-objective optimization algorithms for intrusion detection systems in IoT: a systematic review and future research directions

Article References: Al-Betar, M. A., Mohamed, E. A., & Braik, M. (2026). Multi-objective optimization algorithms for intrusion detection systems in IoT: a systematic review and future research directions. Artificial Intelligence Review. https://doi.org/10.1007/s10462-026-11724-1

Image Credits: AI Generated

DOI: 10.1007/s10462-026-11724-1

Keywords: Internet of Things, intrusion detection systems, multi-objective optimization, network security, swarm intelligence, evolutionary algorithms, machine learning, deep learning, feature selection, false alarm rate, cybersecurity, systematic review

The mathematical skeleton of such a problem is deceptively simple. There is a vector of decision variables, a set of objective functions to be minimized simultaneously, and two families of constraints: inequality constraints that must not be exceeded and equality constraints that must be met exactly. A solution is called feasible if it violates none of the constraints, and the total constraint violation is measured as the sum of the individual violations across every rule. What makes these problems treacherous is that the objectives usually conflict, so no single solution can be best at everything. Instead, the goal is to find a whole set of solutions that are mutually non-dominated, meaning no member of the set is better than another in every objective. This collection of trade-off points, projected into objective space, forms the Pareto front, and when constraints are added, the algorithm must chase the constrained Pareto front, a potentially very different target.

The review highlights a crucial subtlety that often surprises newcomers: the unconstrained Pareto front and the constrained Pareto front can relate to each other in four distinct ways. In the easiest case, the two fronts completely overlap, so every optimal trade-off point is already feasible. In harder cases, the constrained front is a subset of the unconstrained one, or the two only partially overlap, or, most brutally, they are entirely separate, with the unconstrained front lying wholly in infeasible territory. This taxonomy matters because many algorithms work by first pushing a population toward the unconstrained front and then pulling it back toward feasibility. That strategy shines when the two fronts are close, but it can fail badly when they point in different directions, a lesson the authors demonstrate with new experimental comparisons.

At the heart of every constrained multi-objective evolutionary algorithm sit two components: a multi-objective evolutionary engine, which may be dominance-based, decomposition-based, or indicator-based, and a constraint handling technique that decides how to weigh feasible against infeasible solutions during selection. The review walks through the classic techniques with unusual clarity. Penalty function methods fold constraint violation into the fitness value, and the entire art lies in choosing the penalty coefficient: too small and the population never finds feasible solutions, too large and it gets trapped in the first feasible pocket it stumbles into. Static, dynamic, and adaptive variants each try to thread this needle differently.

The constrained domination principle, introduced by Deb and colleagues in the landmark NSGA-II algorithm, takes a more direct route: any feasible solution beats any infeasible one, two infeasible solutions are compared by their degree of violation, and two feasible solutions are compared by ordinary Pareto dominance. It is simple, easy to embed, and converges fast, but its bias toward feasibility makes it hard for a population to cross large infeasible barriers that may separate it from the true constrained front. The epsilon constrained method, due to Takahama and Sakai, softens this by relaxing the constraint boundary with a tolerance parameter, treating mildly infeasible solutions as if they were feasible. As epsilon shrinks toward zero over the run, the method gradually recovers the strict rules while harvesting useful information from high-quality infeasible solutions along the way. Stochastic ranking adds a probabilistic twist, occasionally ignoring constraints entirely with a tunable probability, while multi-objective methods go further and promote constraint violation itself to the status of an extra objective, though this risks inflating the problem into a many-objective one where selection pressure evaporates.

Beyond these foundational techniques, the review organizes the modern algorithm zoo into six families, and this classification is arguably its most useful contribution. Some methods design new fitness functions that blend objectives and constraint violations with adaptive or dynamic weights, often using the proportion of feasible solutions in the population as a feedback signal. Others enhance the classical constraint handling techniques themselves, for example by folding angle information into the constrained domination principle so that high-quality infeasible solutions survive selection and help the population leap across infeasible regions. Constraint relaxation methods tune the epsilon boundary dynamically, sometimes using the maximum and minimum violation values observed among infeasible individuals, and even deploy detect-and-escape mechanisms that relax constraints when the population shows signs of evolutionary stagnation.

Two of the six families have produced some of the field’s most striking recent successes. Two-stage optimization methods split the run into phases: the push-and-pull search framework, for instance, ignores all constraints in the first phase so the population can race to the unconstrained front unimpeded, then uses an improved epsilon method to pull it back toward the constrained front. Auxiliary population methods run a second population alongside the main one, often letting the helper ignore constraints entirely so it can scout the unconstrained front and transfer that knowledge back. The coevolutionary framework CCMO, which pairs a feasibility-focused main population with a constraint-free helper, and the multitasking approach MTCMO, which treats the helper as a separate optimization task with an improved epsilon mechanism, both exemplify this cooperative philosophy. A sixth family rewrites the reproduction operators themselves, using differential evolution mutation strategies that explicitly exploit promising infeasible solutions to generate offspring with better convergence and diversity.

The review does not stop at taxonomy; it also runs head-to-head experiments. Six representative algorithms, including DSPCMDE, MOEADDAE, PPS, CCMO, MTCMO, and CMOCSO, were tested on three benchmark suites with contrasting personalities: LIR-CMOP, whose narrow feasible regions are riddled with infeasible blocks; MW, with tiny, discontinuous feasible regions defined in objective space; and SDC, a newer suite where constraints and objectives share a chaotic, non-monotonic relationship. Each algorithm ran twenty times per problem with a population of one hundred and one hundred thousand evaluations, scored by the inverted generational distance metric on the PlatEMO platform. The results are instructive rather than triumphant for any single method. CMOCSO, with its competitive-and-cooperative swarm operators, dominated LIR-CMOP, taking best or second-best on thirteen of fourteen problems. CCMO and MTCMO led on MW, confirming that mining unconstrained-front information pays off there. But on SDC, where chasing the unconstrained front can actually move the population away from the constrained one, DSPCMDE’s redesigned fitness function proved superior. Friedman’s test ranked CMOCSO first overall and MTCMO second, yet the differences were modest, underscoring the review’s central message that no algorithm wins everywhere and that matching strategy to problem structure is the real game.

The benchmark section itself doubles as a history of the field’s growing rigor, from the pioneering SRN, TNK, and OSY problems of the 1990s through the adjustable-difficulty CTP and DAS-CMOP suites to recent constructions like ZXH_CF, DCMOP with its deceptive constraints that punish solutions precisely where they seem closest to the front, and the real-world RWCMOP collection of fifty engineering problems spanning mechanical design, chemical engineering, power electronics, and power systems. This progression reflects a deliberate effort to expose algorithms to narrow feasible regions, disconnected fronts, multimodality, and deception, features that older, gentler benchmarks simply never tested.

Looking forward, the authors identify five frontiers where current methods strain. Computationally expensive problems, such as those involving computational fluid dynamics, allow only a few thousand real evaluations, making surrogate models that mimic the fitness landscape essential. Multimodal problems hide multiple distinct sets of Pareto-optimal solutions mapping to the same front, demanding strong diversity preservation and mode recognition. Large-scale problems, with hundreds or thousands of decision variables, require dimension reduction and variable grouping to avoid drowning in the search space. Multi-task optimization, which could let experience transfer between related constrained problems, still needs principled knowledge-transfer mechanisms. And dynamic problems, where objectives or constraints shift over time as in fluid catalytic cracking or mineral beneficiation, force algorithms to detect environmental change in real time and decide whether yesterday’s population is worth keeping. With the field already logging thousands of accesses and dozens of citations for this review, its map of what works, what fails, and what remains unsolved is likely to steer the next generation of algorithms toward exactly these open problems.

Subject of Research: Evolutionary algorithms for constrained multi-objective optimization problems

Article Title: Evolutionary constrained multi-objective optimization: a review

Article References: Liang, J., Lin, H., Yue, C., Ban, X., & Yu, K. (2024). Evolutionary constrained multi-objective optimization: a review. Vicinagearth, 1(1), Article 5. https://doi.org/10.1007/s44336-024-00006-5

Image Credits: AI Generated

DOI: 10.1007/s44336-024-00006-5

Keywords: constrained multi-objective optimization, evolutionary algorithms, constraint handling, Pareto front, NSGA-II, epsilon constrained method, benchmark test problems, push and pull search, auxiliary populations, differential evolution, surrogate models, multi-task optimization

The study, published in the International Journal of Machine Learning and Cybernetics by Harshali Zodpe, Shweta Kukade and Manisha Kumawat, introduces the Vehicle Maintenance Optimization Framework, or VMOF. The framework stitches together three distinct computational techniques into a single end-to-end pipeline: Moth-Flame Optimization for feature selection, Differential Evolution for hyperparameter tuning, and an attention-based neural architecture called TabNet for the actual prediction task. The result, according to the authors, is a model that reaches 97.6 percent accuracy with an area under the curve of 0.995 on a real-world vehicle maintenance dataset, comfortably outperforming classical baselines such as support vector machines, decision trees, logistic regression and Naive Bayes.

To understand why this matters, it helps to look at what each component actually does. Predictive maintenance systems are fed streams of sensor readings and operational data, but not all of that information is useful. Irrelevant or redundant features can drown out the genuine warning signals of an impending failure. Moth-Flame Optimization, a nature-inspired algorithm first described in 2015, models the way moths navigate at night by flying at a fixed angle relative to the moon. When the moon is replaced by an artificial flame in the algorithm’s search space, each moth spirals toward the best solutions found so far, and this elegant transverse-orientation maneuver turns out to be remarkably effective at exploring high-dimensional spaces. In VMOF, a binary version of the algorithm decides which features to keep and which to discard, enforcing sparsity so that the final model relies only on the most informative signals.

Feature selection, however, is only half the battle. Every machine learning model carries a set of hyperparameters, settings such as learning rates, regularization strengths and network depths that are not learned from data but must be chosen by the engineer. Getting them wrong can cripple even the best architecture. Rather than relying on grid search or random search, which waste enormous computational resources evaluating poor configurations, VMOF employs Differential Evolution, an evolutionary algorithm that maintains a population of candidate solutions and iteratively improves them by combining existing candidates with weighted differences between others. This continuous optimization step tunes TabNet’s settings automatically, treating the entire maintenance prediction problem as a constrained optimization task whose objective is to minimize classification loss while keeping the feature set lean.

The prediction engine itself, TabNet, is a neural network designed specifically for tabular data, the kind of structured records that dominate industrial datasets. What sets TabNet apart from conventional deep networks is its sequential attention mechanism. At each decision step, the network learns to select which features to attend to, mimicking the way a tree-based model like a random forest splits on the most relevant variables. This gives TabNet two properties that are precious in industrial settings: strong performance on raw tabular data without extensive preprocessing, and a degree of interpretability, because the attention masks reveal which features the model considered important for each individual prediction. For maintenance engineers who need to justify why a vehicle is being pulled off the road, that transparency is not a luxury.

The experimental validation pitted VMOF against a lineup of established classifiers on a real-world vehicle maintenance dataset. The margins were striking. Support vector machines, the strongest classical competitor, managed 90.7 percent accuracy with an AUC of 0.93. Decision trees reached 85.6 percent accuracy with an AUC of 0.87, logistic regression 84.3 percent with an AUC of 0.88, and Naive Bayes trailed at 57 percent accuracy with an AUC of 0.64. VMOF’s 97.6 percent accuracy and 0.995 AUC therefore represent not an incremental improvement but a substantial leap. The framework also achieved a minimum mean squared error of 0.02 and a balanced precision-recall profile, reflected in an F1 score of 0.98, indicating that it is equally adept at catching genuine failures and avoiding false alarms, two goals that often pull in opposite directions.

The balanced F1 score deserves particular attention because predictive maintenance is a domain where the costs of errors are asymmetric. A false positive means a vehicle is unnecessarily serviced, wasting money and downtime. A false negative means a failure occurs without warning, potentially stranding drivers or, in fleet and safety-critical contexts, causing accidents. A model that excels at one while neglecting the other is of limited practical value. By optimizing for classification loss under sparsity constraints and reporting both precision and recall, the authors signal that VMOF is designed for deployment realities rather than leaderboard aesthetics. The interpretability of the attention mechanism further supports this, allowing operators to trace which sensor readings drove a particular maintenance recommendation.

The broader context makes the contribution timely. Research on artificial intelligence-driven vehicle fault diagnosis has accelerated sharply, with recent surveys cataloguing deep learning approaches to remaining useful life prediction, battery health monitoring and Industry 4.0 fault detection. Yet many of these systems share the weaknesses the Pune team set out to fix: pipelines assembled from disconnected components, manual feature engineering that does not transfer between fleets, and hyperparameter searches that consume computing budgets better spent elsewhere. By framing the whole problem as a single constrained optimization task, VMOF offers a template that could, in principle, be adapted beyond automotive maintenance to wind turbines, manufacturing lines and other industrial assets where tabular sensor data and scarce expert attention collide.

There are, of course, caveats. The authors report that no new datasets were generated or analyzed during the study beyond the benchmark data used for validation, and the framework’s performance will ultimately need to be demonstrated across diverse vehicle types, sensor configurations and operating conditions before it becomes a standard tool in service garages or fleet management platforms. Evolutionary optimization, while powerful, adds its own computational overhead, and the interplay between the moth-inspired search and differential evolution would need careful calibration for each new deployment. The study also arrives at a moment when the field is still grappling with class imbalance, a problem highlighted in related work on automotive maintenance prediction, where failure events are by definition rare compared with normal operation.

Even so, the study’s central message is likely to resonate well beyond the automotive sector. It demonstrates that combining evolutionary search with sparse, attention-based learning can simultaneously improve accuracy, interpretability and deployability, three qualities that are usually treated as trade-offs. A moth’s spiral toward a distant flame is, on the face of it, a strange metaphor for machine reliability. But if the numbers hold, the navigation trick that guides moths through the night may help keep vehicles on the road, turning maintenance from a reactive scramble into a precise, data-driven science.

Subject of Research: Machine learning-based predictive maintenance for vehicles using evolutionary feature selection and attention-based tabular deep learning

Article Title: Optimizing predictive maintenance in the automotive sector with Moth-Flame Optimization and TabNet

Article References: Zodpe, H., Kukade, S., & Kumawat, M. (2026). Optimizing predictive maintenance in the automotive sector with Moth-Flame Optimization and TabNet. International Journal of Machine Learning and Cybernetics, 17(10), Article 483. https://doi.org/10.1007/s13042-026-03322-y

Image Credits: AI Generated

DOI: 10.1007/s13042-026-03322-y

Keywords: predictive maintenance, automotive industry, Moth-Flame Optimization, Differential Evolution, TabNet, feature selection, hyperparameter optimization, machine learning, attention mechanism, vehicle fault prediction, interpretability, evolutionary algorithms