Maintaining Plasticity in Neural Networks

Nonstationarity presents a variety of challenges for machine learning systems. One surprising pathology which can arise in nonstationary learning problems is plasticity loss, whereby making progress on new learning objectives becomes more difficult as training progresses. Networks which are unable to adapt in response to changes in their environment experience plateaus or even declines in performance in highly non-stationary domains such as reinforcement learning, where the learner must quickly adapt to new information even after hundreds of millions of optimization steps. The loss of plasticity manifests in a cluster of related empirical phenomena which have been identified by a number of recent works, including the primacy bias, implicit under-parameterization, rank collapse, and capacity loss. While this phenomenon is widely observed, it is still not fully understood. This talk will present exciting recent results which shed light on the mechanisms driving the loss of plasticity in a variety of learning problems and survey methods to maintain network plasticity in non-stationary tasks, with a particular focus on deep reinforcement learning.

Relations and Predictions in Brains and Machines

Humans and animals learn and plan with flexibility and efficiency well beyond that of modern Machine Learning methods. This is hypothesized to owe in part to the ability of animals to build structured representations of their environments, and modulate these representations to rapidly adapt to new settings. In the first part of this talk, I will discuss theoretical work describing how learned representations in hippocampus enable rapid adaptation to new goals by learning predictive representations, while entorhinal cortex compresses these predictive representations with spectral methods that support smooth generalization among related states. I will also cover recent work extending this account, in which we show how the predictive model can be adapted to the probabilistic setting to describe a broader array of generalization results in humans and animals, and how entorhinal representations can be modulated to support sample generation optimized for different behavioral states. In the second part of the talk, I will overview some of the ways in which we have combined many of the same mathematical concepts with state-of-the-art deep learning methods to improve efficiency and performance in machine learning applications like physical simulation, relational reasoning, and design.

The power of structured representations (and how to learn them)

UCL NeuroAI annual half day event (hybrid)

Towards a recipe for physical reasoning in humans and machines

Transforming task representations

Humans can adapt to a novel task on our first try. By contrast, artificial intelligence systems often require immense amounts of data to adapt. In this talk, I will discuss my recent work (https://www.pnas.org/content/117/52/32970) on creating deep learning systems that can adapt on their first try by exploiting relationships between tasks. Specifically, the approach is based on transforming a representation for a known task to produce a representation for the novel task, by inferring and then using a higher order function that captures a relationship between the tasks. This approach can be interpreted as a type of analogical reasoning. I will show that task transformation can allow systems to adapt to novel tasks on their first try in domains ranging from card games, to mathematical objects, to image classification and reinforcement learning. I will discuss the analogical interpretation of this approach, an analogy between levels of abstraction within the model architecture that I refer to as homoiconicity, and what this work might suggest about using deep-learning models to infer analogies more generally.

Mental Simulation, Imagination, and Model-Based Deep RL

Mental simulation—the capacity to imagine what will or what could be—is a salient feature of human cognition, playing a key role in a wide range of cognitive abilities. In artificial intelligence, the last few years have seen the development of methods which are analogous to mental models and mental simulation. In this talk, I will discuss recent methods in deep learning for constructing such models from data and learning to use them via reinforcement learning, and compare such approaches to human mental simulation. While a number of challenges remain in matching the capacity of human mental simulation, I will highlight some recent progress on developing more compositional and efficient model-based algorithms through the use of graph neural networks and tree search.

Graph Representation Learning and the Hippocampal-Entorhinal Circuit

The Learning Salon

In the Learning Salon, we will discuss the similarities and differences between biological and machine learning, including individuals with diverse perspectives and backgrounds, so we can all learn from one another.

The Spatial Memory Pipeline: a deep learning model of egocentric to allocentric understanding in mammalian brains

The Learning Salon

In the Learning Salon, we will discuss the similarities and differences between biological and machine learning, including individuals with diverse perspectives and backgrounds, so we can all learn from one another.

Understanding sensorimotor control at global and local scales

The brain is remarkably flexible, and appears to instantly reconfigure its processing depending on what’s needed to solve a task at hand: fMRI studies indicate that distal brain areas appear to fluidly couple and decouple with one another depending on behavioral context. We investigated how the brain coordinates its activity across areas to inform complex, top-down control behaviors. Animals were trained to perform a novel brain machine interface task to guide a visual cursor to a reward zone, using activity recorded with widefield calcium imaging. This allowed us to screen for cortical areas implicated in causal neural control of the visual object. Animals could decorrelate normally highly-correlated areas to perform the task, and used an explore-exploit search in neural activity space to discover successful strategies. Higher visual and parietal areas were more active during the task in expert animals. Single unit recordings targeted to these areas indicated that the sensory representation of an object was sensitive to an animal’s subjective sense of controlling it.