Researchers led by Omid Farhangi and Ahmad Banakar of Tarbiat Modares University, together with colleagues in Iran, set out to replace the slow, destructive laboratory tests currently used to detect PPD with a fully non-destructive optical method. Conventional approaches rely on high-performance liquid chromatography or gas chromatography–mass spectrometry, which deliver excellent accuracy but demand lengthy sample preparation, hazardous solvents, expensive instruments, and trained specialists. For market surveillance, customs checkpoints, and factory quality control, that combination is impractical. The team’s alternative was to shine ultraviolet, visible, and near-infrared light onto henna powder, record the reflected spectrum, and let algorithms translate the optical fingerprint into an exact PPD concentration.

The experimental design was deliberately rigorous. Pure henna leaves were collected from the major cultivation regions of Kerman and Sistan-Baluchestan in Iran, ground under controlled laboratory conditions, and supplemented with market samples from Yazd province. Laboratory-grade PPD was then mixed into the powder at seven levels ranging from 0 to 12 percent by weight, a range chosen to mirror the contamination levels actually reported in adulterated commercial henna. Eleven independently prepared samples were made at each concentration, yielding 77 physical samples in total. From each one, five reflectance spectra were recorded at different spots and averaged before any data splitting, a crucial step that prevented replicate measurements from leaking information between training and test sets.

The spectroscopy itself relied on an Ocean Optics USB2000 spectrometer covering 179 to 872 nanometers, illuminated by a UV-B lamp for the ultraviolet region and a halogen source for the visible and near-infrared. Light was delivered and collected through a bifurcated fiber probe, with samples held at a fixed thickness inside a black box to exclude ambient light. The raw spectra revealed a striking feature: a sharp, strong reflectance peak at 380 to 390 nanometers that grew steadily in intensity as PPD concentration increased. This peak arises from the conjugated pi-electron system of PPD’s aromatic amine structure, which interacts strongly with ultraviolet radiation, whereas lawsone’s quinonoid structure responds far more weakly in this window.

Principal component analysis confirmed what the spectra suggested. In the UV region, samples clustered cleanly by PPD concentration along the first principal component, demonstrating that the optical features were chemically specific and systematically tied to contamination. In the visible and near-infrared region, by contrast, no meaningful clustering appeared; the samples lined up along a single trend driven by overall reflectance offset, a physical scattering effect rather than a chemical signature. This distinction proved predictive of the modeling results: the UV region would reward nonlinear machine learning, while the visible region would struggle across the board.

The team compared five regression algorithms: partial least squares regression, decision tree, random forest, support vector regression, and an artificial neural network. On full-spectrum UV data, the decision tree with no preprocessing achieved a test coefficient of determination of 0.9376 with a residual prediction deviation of 4.14, edging out PLSR’s 0.9339. Random forest and support vector regression also performed respectably. In the visible region, performance dropped sharply, with the best model, PLSR with detrending, reaching only 0.7583 on the test set, still above the conventional threshold of 2.0 RPD that indicates a very good model but far below UV-region accuracy.

The study’s most innovative stage involved six metaheuristic optimization algorithms, including genetic algorithms, particle swarm optimization, ant colony optimization, and learning automata, each tasked with selecting the 15 most informative wavelengths from hundreds of spectral variables. Learning automata emerged as the clear winner in both spectral regions, achieving the lowest error and highest correlation in the UV range. Its advantage stems from an adaptive probability-update mechanism that rewards useful wavelengths and punishes uninformative ones, allowing it to escape local optima that trapped faster competitors. The genetic algorithm, despite the shortest runtime of nine seconds, performed worst in the UV region, highlighting the trade-off between computational speed and convergence quality.

Feature selection delivered a dramatic payoff. After reducing the data to just 15 wavelengths, more than 90 percent of the original spectral variables discarded, random forest with Gaussian filter preprocessing leapt from a test R-squared of 0.9186 to 0.9779, with an outstanding RPD of 6.99, the best result of the entire study. The artificial neural network followed closely with an R-squared of 0.9476 and RPD of 4.51. The improvement shows that ensemble methods, often destabilized by noisy and collinear spectral variables, can unlock their full potential when fed only the most diagnostic features. Interestingly, the learning automata selected wavelengths above 400 nanometers in addition to the characteristic UV peak, indicating that scattering-related changes in the visible range carry complementary information about PPD content.

The practical implications extend well beyond henna. Because the entire analysis depends on only 15 key wavelengths, manufacturers could build dedicated, low-cost multispectral sensors that operate at those specific optical points, deploying them on production lines, at customs checkpoints, and in retail settings for real-time screening. Such devices would align with the broader trend toward miniaturized analytical systems in the food and cosmetic industries. The authors caution, however, that their dataset was generated under controlled laboratory conditions, and real commercial samples may contain interfering compounds, variable particle sizes, and moisture that could affect spectral responses. Validation against chromatographic reference methods on diverse commercial products, calibration transfer protocols for portable instruments, and cross-dataset testing of the selected wavelengths remain necessary next steps.

The work also opens the door to broader applications of the same framework. Future research could extend the approach to simultaneous detection of multiple adulterants, including other aromatic amines, heavy metals, and plant-based fillers, or employ deep learning architectures such as one-dimensional convolutional neural networks and data fusion strategies combining UV-Vis-NIR spectroscopy with Raman techniques. For now, the study stands as a compelling demonstration that an ancient cosmetic product can be protected by thoroughly modern tools: a beam of light, a fiber-optic probe, and algorithms smart enough to read the chemical secrets hidden in a reflection. In a market where a single contaminated tattoo can cause lifelong injury, that combination may prove invaluable.

Subject of Research: Non-destructive detection of para-phenylenediamine adulteration in henna using UV–Vis–NIR spectroscopy and machine learning

Article Title: Non-destructive quantification of para-phenylenediamine in henna using UV–Vis–NIR spectroscopy and machine learning

Article References: Farhangi, O., Banakar, A., Ebadi, M.-T., & Mahdavian, A. (2026). Non-destructive quantification of para-phenylenediamine in henna using UV–Vis–NIR spectroscopy and machine learning. Smart Agricultural Technology, 15, Article 102621. https://doi.org/10.1016/j.atech.2026.102621

Image Credits: AI Generated

DOI: Not provided

Keywords: henna, para-phenylenediamine, UV-Vis-NIR spectroscopy, machine learning, food adulteration, cosmetic safety, chemometrics, wavelength selection, learning automata, random forest, artificial neural network, quality control

The system, called IGDHRS for intelligent graph-based deep hybrid recommender system, was developed by Milad Payandeh, Seyed Mahdi Jameii, and Mostafa Haghi Kashani of the Department of Computer Engineering at Islamic Azad University in Iran. Their starting point is a familiar taxonomy: recommender systems generally fall into collaborative filtering, which learns from patterns in user behavior; content-based filtering, which matches item attributes to user preferences; and hybrid models that blend the two. Each family carries its own liabilities. Collaborative approaches collapse when interaction data is thin or missing, while content-based methods struggle to capture the subtle, evolving tastes that ratings reveal. The Iranian team’s answer is a hybrid that treats the user population itself as a graph and then learns rich representations from that structure.

The first architectural step is the construction of a user–user similarity graph, in which nodes represent individual users and edges encode how alike their rating behaviors are. Building such a graph requires deciding, for every pair of users, whether their measured similarity is strong enough to justify a connection, and that decision hinges on a similarity threshold. Set the threshold too low and the graph becomes a dense tangle of weakly related users, diluting the signal; set it too high and the graph fragments, cutting off genuinely helpful neighborhood information. Rather than fixing this threshold by hand, the researchers let it adapt dynamically using learning automata, a class of reinforcement-driven stochastic decision units that adjust their actions based on feedback from the environment. In effect, the system tunes its own notion of who counts as a similar user as training proceeds.

Learning automata deserve a closer look because they are a departure from the gradient-based optimization that dominates modern deep learning. An automaton maintains a probability distribution over a set of possible actions, selects one, observes a reward or penalty, and updates its probabilities accordingly. Over many iterations it converges toward actions that consistently earn rewards. In IGDHRS, that feedback loop nudges the similarity thresholds toward values that ultimately improve recommendation quality, a form of automatic hyperparameter adaptation that relieves engineers of a delicate tuning burden and lets the graph topology evolve to match the data at hand.

To give the graph more expressive power, the authors enrich each user node with auxiliary demographic information, including age, gender, and occupation. This is a deliberate countermeasure to the cold-start problem: even a brand-new user with zero ratings carries demographic attributes that can anchor them in the similarity graph, linking them to established users with comparable profiles. Previous work has shown that demographic profile expansion can buffer sparse rating matrices, but integrating such side information directly into a graph neural architecture is what makes this system distinctive. The demographics act as a bridge across the rating desert, allowing information to propagate from well-modeled users to newcomers along graph edges that would not otherwise exist.

At the heart of the architecture sits the study’s central technical contribution: a deep denoising graph convolutional autoencoder, abbreviated DDGCAE. An autoencoder is a neural network trained to compress its input into a low-dimensional latent code and then reconstruct the original signal from that code, forcing it to learn the essential structure of the data. A denoising autoencoder raises the stakes by deliberately corrupting the input, for example by masking or perturbing entries, and requiring the network to recover the clean version, which cultivates robustness to the missing and noisy values that pervade real rating matrices. A graph convolutional autoencoder extends this idea to graph-structured data: graph convolution layers aggregate information from each node’s neighbors, so the learned embeddings encode not just a user’s own behavior but the behavior of the surrounding network neighborhood.

Combining all three ingredients means that DDGCAE operates on an enriched, adaptively thresholded user similarity graph, learns compressed representations by reconstructing denoised graph signals, and produces latent user profiles that capture both interaction patterns and demographic context. Those latent representations then drive rating prediction. The design echoes and extends a lineage of prior systems, from classic autoencoder-based collaborative filtering models such as AutoRec and collaborative denoising autoencoders to graph-based methods like Neural Graph Collaborative Filtering and LightGCN, but the authors argue that the joint interplay of denoising, graph convolution, and automata-driven graph construction is what sets their approach apart.

The team implemented IGDHRS in Python and evaluated it on four widely used benchmark datasets spanning different scales and domains: MovieLens 100K, MovieLens 1M, a stratified random sample of the Netflix Prize data, and the Amazon Movies and TV dataset. Because the Netflix and Amazon corpora are enormous, the researchers extracted stratified random samples, and they have made the sampling scripts and generated sample indices publicly available in a GitHub repository, alongside the full system implementation, a level of openness that supports reproducibility. Performance was measured with standard regression and ranking metrics: root mean squared error and mean absolute error to quantify how far predictions deviate from true ratings, and precision and recall to gauge the quality of the top recommendations actually surfaced to users.

The reported results are striking. Across all four datasets, the proposed system significantly outperformed several state-of-the-art comparison methods, and the advantages held in the conditions that matter most: domains with higher data sparsity and datasets where demographic information was partially missing. The authors attribute this robustness directly to the three-way combination of auxiliary user data, learning automata, and the DDGCAE architecture, and they specifically highlight improved resilience to cold-start situations, the scenario in which traditional collaborative filtering degrades most severely. Statistical rigor was addressed as well, with the study employing cross-validation practices and nonparametric significance testing in the tradition of the Wilcoxon ranking method to substantiate that observed gains were not artifacts of a lucky split.

For the broader field, the study suggests that the path past the cold-start and sparsity bottleneck may lie not in any single clever component but in architectures that let multiple adaptive mechanisms reinforce one another. Graph structures supply the relational scaffolding, denoising objectives harden the learned embeddings against missing data, demographic side channels keep new users connected from day one, and learning automata quietly optimize the structural choices that humans would otherwise guess. The code and data are public, the benchmarks are the community’s standards, and the message is clear: recommender systems that can rebuild their own wiring while learning from corrupted signals are a promising blueprint for the next generation of personalization engines, from streaming catalogs to e-commerce and beyond.

Subject of Research: A deep graph-based hybrid recommender system addressing cold-start and data sparsity using learning automata

Article Title: An intelligent recommender system based on deep denoising graph convolutional autoencoder and learning automata

Article References: Payandeh, M., Jameii, S. M., & Kashani, M. H. (2026). An intelligent recommender system based on deep denoising graph convolutional autoencoder and learning automata. International Journal of Data Science and Analytics, 22(1), Article 312. https://doi.org/10.1007/s41060-026-01293-5

Image Credits: AI Generated

DOI: 10.1007/s41060-026-01293-5

Keywords: recommender systems, graph convolutional networks, autoencoders, learning automata, cold-start problem, data sparsity, collaborative filtering, deep learning, MovieLens, Netflix Prize, Amazon dataset, hybrid recommendation