Their creation, called Dynamic Continuous Progressive Neural Networks, or DYNcPNN, addresses a scenario the authors term Evolving Streaming Time Series. In such settings, data arrives as an unbounded stream, past values influence future ones, the underlying statistical relationships shift without warning, and the model must somehow absorb new knowledge without erasing what it has already learned. Each of these demands has traditionally been the province of a separate research field. Streaming Machine Learning emphasizes rapid adaptation to drifts but happily abandons old knowledge in the process. Continual Learning protects deep networks from catastrophic forgetting but assumes drifts merely introduce new data distributions rather than contradictory ones. Time Series Analysis, particularly through recurrent architectures like Long Short-Term Memory networks, handles temporal dependence but is rarely deployed in true streaming contexts. The Milanese team’s Streaming Continual Learning paradigm, introduced in previous work and now embodied in DYNcPNN, unifies all three.

The problem DYNcPNN confronts is subtler than it first appears. Concept drift comes in flavors that demand different responses. The authors draw a novel distinction between what they call contradictory drifts and input drifts. A contradictory drift changes the decision boundary itself: the same input that once meant “normal conditions” may now signify a storm, perhaps because an agency has revised its alert thresholds. An input drift, by contrast, introduces unfamiliar data—new slang in customer reviews, novel sensor readings—without invalidating the old classification rules. The difference matters enormously. Contradictory drifts force a model to relearn, while input drifts reward a model that can weave new knowledge into existing structures without disturbing them. Real streams present both, often at once, along with temporal dependence, where the probability of an event at time t genuinely depends on what happened at t minus some lag. Standard streaming classifiers, which typically assume data points are independent, ignore this temporal structure entirely and pay for it in accuracy.

DYNcPNN builds on the team’s earlier architecture, cPNN, which itself adapted the Progressive Neural Networks strategy from Continual Learning to the streaming domain. The original Progressive Neural Networks grow by adding an entirely new “column” of neural network layers whenever a new task appears, while freezing the weights of previous columns so their knowledge remains intact. Transfer connections let new columns draw on old columns’ representations, enabling selective reuse of past expertise. The cPNN variant applied this logic on top of a Continuous LSTM—a recurrent base learner that buffers the incoming stream in fixed-size mini-batches and builds sequences with a hopping window—thereby capturing temporal dependence while learning continuously. But cPNN suffered from two crippling limitations: it required knowing the exact timestamps of concept drifts in advance, and it expanded its architecture every single time a drift was detected, whether or not such expansion was warranted.

The new work eliminates both limitations through two intertwined mechanisms. The first is an automatic concept drift detector based on ADWIN—ADaptive WINdowing—a well-established streaming technique that maintains a sliding window of recent observations, here a binary correctness signal of the model’s predictions. ADWIN continuously evaluates all ways of splitting the window into a recent half and an older half; when the difference between the means of the two sub-windows exceeds a statistically motivated threshold, a drift is signaled and the oldest observations are discarded. By wiring this detector into the architecture, DYNcPNN no longer needs to be told when the world has changed—it notices on its own, without manual intervention.

The second and arguably more consequential innovation is the dynamic decision mechanism that governs when the network actually expands. When a drift is detected, expanding the architecture is not always the right response. Adding a column costs memory, and if the new concept is mild or closely related to the current one, simple continued training on the existing network may suffice. DYNcPNN therefore evaluates whether the drift is severe enough to justify growth. If the drift is judged contradictory and substantial, a new column is added and the old ones are frozen, preserving their knowledge forever. If, however, the model decides to adapt without expanding—continuing to train the existing network—it risks overwriting previously learned knowledge. To prevent exactly this form of catastrophic forgetting, the authors introduce a protective strategy that safeguards the relevant prior knowledge during this no-expansion adaptation phase. The net effect is a model that spends memory only when necessary and consistently outperforms its predecessors in both accuracy and footprint.

The severity of a drift, which drives this decision, is formalized in the paper as the discrepancy between the joint probability distributions before and after the change. Mild drifts require only minor adjustments; severe ones demand substantial relearning. For contradictory drifts, severity can be gauged through the “influence zone”—the proportion of the previously observed input space where the new concept changed the labels. This quantitative framing allows the architecture to make principled rather than arbitrary choices about when to grow.

The experimental campaign is notably thorough. The team benchmarked DYNcPNN against its own predecessor cPNN, against the underlying cLSTM trained continuously, and against the heavyweights of the Streaming Machine Learning world, including Hoeffding Adaptive Trees and Adaptive Random Forests. The benchmarks were crafted to include significant temporal dependence: the authors injected elaborate temporal dependencies into the labels of synthetic streams, and they also tested on real-world data covering weather prediction, air pollution, and power consumption—domains where sensor readings evolve continuously and past measurements genuinely shape the future. The results, reported in the paper, show DYNcPNN consistently outperforming all competitors, adapting more quickly to concept drifts, mitigating catastrophic forgetting effectively, and optimizing memory usage. Equally revealing were the failures of the baselines: the streaming decision-tree and forest models, unable to account for temporal dependence, lagged behind on all temporally structured tasks.

The significance of this work extends beyond a single architecture. It serves as a pointed critique of the assumptions underpinning much of classical machine learning. Offline learning rests on the i.i.d. assumption—that data are independent and identically distributed—which the Empirical Risk Minimization principle requires for meaningful statistical guarantees. Streaming data with temporal dependence and concept drift violates both conditions at once. Correlated samples reduce the effective number of independent observations, biasing risk estimates, while shifting distributions make convergence impossible in principle. The authors’ response is not to patch the old framework but to build one designed for the violation, drawing on an emerging body of SCL research that has recently begun to emphasize taming temporal dependence in streams.

Practical applications are easy to imagine. The paper’s own motivating example is environmental nowcasting: a classifier receiving an unbounded stream of temperature, humidity, and radiation measurements from sensors must predict the current weather condition in real time, before official meteorological confirmation arrives. Seasonal transitions produce input drifts; revised alert thresholds produce contradictory drifts; rain persistence and gradual temperature trends produce temporal dependence; and recurring phenomena—summer storms, winter snow showers—demand that old knowledge remain retrievable rather than destroyed. The same structure applies to network security, financial monitoring, industrial sensor networks, and any Internet of Things deployment where the ground truth arrives late or never.

The authors are candid about limits and future directions, and DYNcPNN is presented as a pioneering embodiment of a young paradigm rather than a finished solution. Yet the trajectory is clear: as data-hungry applications proliferate and the gap between laboratory conditions and operational reality widens, architectures that learn continuously, remember selectively, and adapt autonomously will become less exotic and more essential. With DYNcPNN, the Streaming Continual Learning paradigm has gained a concrete, benchmarked, and open blueprint for what that future might look like—neural networks that grow only when they must, forget only what they should, and never stop reading the stream.

Research has demonstrated that social species—those organisms that thrive within intricate interpersonal networks—exhibit behavioral adaptations based on real-time interactions with peers. These adaptations are imperative as the rewards and penalties associated with various behaviors can fluctuate. For example, in a group of social animals, an observed behavior may yield different outcomes based on immediate context or the behavior of other individuals in the group. This fluidity in behavioral strategy highlights a profound layer of complexity that is often missing in traditional AI frameworks, which do not typically adjust or learn after their initial training phase.

Neuroscience offers a wealth of information regarding how living organisms navigate and adapt their behaviors in response to changes in their environment. An extensive body of research outlines how animals learn in conditions where rules, reward structures, and expected outcomes are neither fixed nor predictable. This contrasts starkly with the conventional training paradigm of AI systems, which are often siloed in their learning strategies, unable to evolve after deployment. As AI technologies advance and begin to intertwine with the fabric of daily life—guiding everything from autonomous vehicles to personal assistants—there is a compelling imperative to re-evaluate the rigidity of these systems through a neuroscientific lens.

Consider the intricacies of how animals learn within social environments. For instance, studies on primates and other social mammals reveal that group dynamics can instigate shifts in learning behaviors, adapting strategies based on the behaviors of others and altering outcomes in real-time. Such observations provide valuable insights into the malleability of learning. By leveraging principles derived from neuroscience, AI could potentially be engineered to adjust dynamically to such nuanced scenarios, paving the way for more intelligent and responsive systems capable of real-world application.

The concept of continual learning—iterations of learning that adapt over time—has gained traction in recent AI research. However, the prevailing models typically struggle with ‘catastrophic forgetting,’ a phenomenon where the introduction of new information leads to the deterioration of previously acquired knowledge. Utilizing insights from neuroscience could provide strategies to overcome these limitations, enhancing AI systems’ abilities to retain learned information while also adapting to new data. The dynamic nature of animal learning can inspire frameworks that allow AI to develop more resilient architectures better equipped for real-world applications.

The mechanisms through which animals encode and recall information about their environment are profound areas of inquiry within neuroscience. Recent studies on neural activity in various species illustrate the way neuronal populations transition rapidly in response to changing inputs, demonstrating the brain’s ability to encode multidimensional tasks efficiently. Emulating these neural mechanisms could greatly enhance machine-learning algorithms, allowing them to process changing information in a manner akin to how biological entities operate.

Moreover, behavioral experiments indicate that animals often engage in explorative behavior to ascertain new rules or rewards within their environment, a process that is facilitated by neuroplasticity—the brain’s ability to reorganize itself by forming new neural connections. Such exploration strategies can inform AI frameworks that would encourage agents to seek out novel data and experiences actively, fostering an environment of continuous learning and adaptation. By merging neuroscience principles with AI architectures, the path towards developing more autonomous and capable systems opens up, effectively bridging the gap between artificial learning and natural intelligence.

In pursuit of this interdisciplinary dialogue, researchers are called upon to integrate established knowledge from neuroscience into the evolving field of AI, promoting mutual understanding and innovation. Collaborative efforts can lead to rich exchanges of ideas that bolster the growth of both domains, creating a synergy that advances our comprehension of learning and adaptability. AI systems robust enough to mimic animal-like learning capabilities could positively impact various applications, including robotics, healthcare, and user interface design.

As advancements in AI continue to accelerate, the necessity for systems characterized by adaptive learning becomes paramount. The ability to adjust based on tangible experiences and interactions not only enhances performance but also has profound ethical implications as AI begins to operate in sensitive domains such as healthcare or security. By understanding how animals regulate behavior based on social context and environmental feedback, AI researchers can better equip systems for ethical reasoning and decision-making in the complex tapestry of human interaction.

The exploration of how neuroscience can inform the development of AI serves as an intriguing frontier for investigation, pushing the boundaries of our understanding of both realms. As such, there is an urgent call to further this research agenda, creating robust collaborative frameworks that would allow scientists, engineers, and ethicists to work together in conceiving AI systems that learn and adapt continuously. This endeavor is not merely an academic pursuit; it embodies our intrinsic desire to understand intelligence—both artificial and natural—ultimately enriching our grasp over the technologies that shape our future.

As the discourse surrounding the intersection of AI and neuroscience grows, it will undoubtedly illuminate new pathways towards intelligent systems that are more nuanced, adaptable, and intertwined with human experiences. The journey involves embracing complexity, allowing AI to learn not just through vast data collections but through intelligent interaction, mirroring the remarkable learning capabilities seen in the animal kingdom. As we navigate this interdisciplinary convergence, we stand at the threshold of a revolution in how technology can evolve, potentially transforming the fabric of our relationship with machines and redefining our expectations of intelligent behavior.

Subject of Research: The intersection of neuroscience and artificial intelligence in understanding learning in dynamic environments.

Article Title: What neuroscience can tell AI about learning in continuously changing environments.

Article References:

Durstewitz, D., Averbeck, B. & Koppe, G. What neuroscience can tell AI about learning in continuously changing environments.
Nat Mach Intell (2025). https://doi.org/10.1038/s42256-025-01146-z

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s42256-025-01146-z

Keywords: NeuroAI, artificial intelligence, neuroscience, continual learning, adaptive systems, dynamic environments