arXiv subjects

Samuel J. Gershman

Publications and source records attributed to Samuel J. Gershman.

At least 37 records · Page 2Linked to original sources

arXiv2025

Synthesizing world models for bilevel planning

Modern reinforcement learning (RL) systems have demonstrated remarkable capabilities in complex environments, such as video games. However, they still fall short of achieving human-like sample efficiency and adaptability when learning new domains. Theory-based reinforcement learning (TBRL) is an algorithmic framework specifically designed to address this gap. Modeled on cognitive theories, TBRL leverages structured, causal world models - "theories" - as forward simulators for use in planning, generalization and exploration. Although current TBRL systems provide compelling explanations of how humans learn to play video games, they face several technical limitations: their theory languages are restrictive, and their planning algorithms are not scalable. To address these challenges, we introduce TheoryCoder, an instantiation of TBRL that exploits hierarchical representations of theories and efficient program synthesis methods for more powerful learning and planning. TheoryCoder equips agents with general-purpose abstractions (e.g., "move to"), which are then grounded in a particular environment by learning a low-level transition model (a Python program synthesized from observations by a large language model). A bilevel planning algorithm can exploit this hierarchical structure to solve large domains. We demonstrate that this approach can be successfully applied to diverse and challenging grid-world games, where approaches based on directly synthesizing a policy perform poorly. Ablation studies demonstrate the benefits of using hierarchical abstractions.

cs.AI↗

arXiv2025

Adaptive Social Learning using Theory of Mind

Social learning is a powerful mechanism through which agents learn about the world from others. However, humans don't always choose to observe others, since social learning can carry time and cognitive resource costs. How do people balance social and non-social learning? In this paper, we propose a rational mentalizing model of the decision to engage in social learning. This model estimates the utility of social learning by reasoning about the other agent's goal and the informativity of their future actions. It then weighs the utility of social learning against the utility of self-exploration (non-social learning). Using a multi-player treasure hunt game, we show that our model can quantitatively capture human trade-offs between social and non-social learning. Furthermore, our results indicate that these two components allow agents to flexibly apply social learning to achieve their goals more efficiently.

cs.MA↗

arXiv2025

Key-value memory in the brain

Classical models of memory in psychology and neuroscience rely on similarity-based retrieval of stored patterns, where similarity is a function of retrieval cues and the stored patterns. While parsimonious, these models do not allow distinct representations for storage and retrieval, despite their distinct computational demands. Key-value memory systems, in contrast, distinguish representations used for storage (values) and those used for retrieval (keys). This allows key-value memory systems to optimize simultaneously for fidelity in storage and discriminability in retrieval. We review the computational foundations of key-value memory, its role in modern machine learning systems, related ideas from psychology and neuroscience, applications to a number of empirical puzzles, and possible biological implementations.

q-bio.NC↗

arXiv2024

Artificial intelligence for science: The easy and hard problems

A suite of impressive scientific discoveries have been driven by recent advances in artificial intelligence. These almost all result from training flexible algorithms to solve difficult optimization problems specified in advance by teams of domain scientists and engineers with access to large amounts of data. Although extremely useful, this kind of problem solving only corresponds to one part of science - the "easy problem." The other part of scientific research is coming up with the problem itself - the "hard problem." Solving the hard problem is beyond the capacities of current algorithms for scientific discovery because it requires continual conceptual revision based on poorly defined constraints. We can make progress on understanding how humans solve the hard problem by studying the cognitive science of scientists, and then use the results to design new computational agents that automatically infer and update their scientific paradigms.

cs.AI↗

arXiv2024

Do Mice Grok? Glimpses of Hidden Progress During Overtraining in Sensory Cortex

Does learning of task-relevant representations stop when behavior stops changing? Motivated by recent theoretical advances in machine learning and the intuitive observation that human experts continue to learn from practice even after mastery, we hypothesize that task-specific representation learning can continue, even when behavior plateaus. In a novel reanalysis of recently published neural data, we find evidence for such learning in posterior piriform cortex of mice following continued training on a task, long after behavior saturates at near-ceiling performance ("overtraining"). This learning is marked by an increase in decoding accuracy from piriform neural populations and improved performance on held-out generalization tests. We demonstrate that class representations in cortex continue to separate during overtraining, so that examples that were incorrectly classified at the beginning of overtraining can abruptly be correctly classified later on, despite no changes in behavior during that time. We hypothesize this hidden yet rich learning takes the form of approximate margin maximization; we validate this and other predictions in the neural data, as well as build and interpret a simple synthetic model that recapitulates these phenomena. We conclude by showing how this model of late-time feature learning implies an explanation for the empirical puzzle of overtraining reversal in animal learning, where task-specific representations are more robust to particular task changes because the learned features can be reused.

cs.LG↗

arXiv2024

Predictive representations: building blocks of intelligence

Adaptive behavior often requires predicting future events. The theory of reinforcement learning prescribes what kinds of predictive representations are useful and how to compute them. This paper integrates these theoretical ideas with work on cognition and neuroscience. We pay special attention to the successor representation (SR) and its generalizations, which have been widely applied both as engineering tools and models of brain function. This convergence suggests that particular kinds of predictive representations may function as versatile building blocks of intelligence.

cs.AI↗

arXiv2024

Reconciling Shared versus Context-Specific Information in a Neural Network Model of Latent Causes

It has been proposed that, when processing a stream of events, humans divide their experiences in terms of inferred latent causes (LCs) to support context-dependent learning. However, when shared structure is present across contexts, it is still unclear how the "splitting" of LCs and learning of shared structure can be simultaneously achieved. Here, we present the Latent Cause Network (LCNet), a neural network model of LC inference. Through learning, it naturally stores structure that is shared across tasks in the network weights. Additionally, it represents context-specific structure using a context module, controlled by a Bayesian nonparametric inference algorithm, which assigns a unique context vector for each inferred LC. Across three simulations, we found that LCNet could 1) extract shared structure across LCs in a function learning task while avoiding catastrophic interference, 2) capture human data on curriculum effects in schema learning, and 3) infer the underlying event structure when processing naturalistic videos of daily events. Overall, these results demonstrate a computationally feasible approach to reconciling shared structure and context-specific structure in a model of LCs that is scalable from laboratory experiment settings to naturalistic settings.

q-bio.NC↗

arXiv2024

Grokking as the Transition from Lazy to Rich Training Dynamics

We propose that the grokking phenomenon, where the train loss of a neural network decreases much earlier than its test loss, can arise due to a neural network transitioning from lazy training dynamics to a rich, feature learning regime. To illustrate this mechanism, we study the simple setting of vanilla gradient descent on a polynomial regression problem with a two layer neural network which exhibits grokking without regularization in a way that cannot be explained by existing theories. We identify sufficient statistics for the test loss of such a network, and tracking these over training reveals that grokking arises in this setting when the network first attempts to fit a kernel regression solution with its initial features, followed by late-time feature learning where a generalizing solution is identified after train loss is already low. We find that the key determinants of grokking are the rate of feature learning -- which can be controlled precisely by parameters that scale the network output -- and the alignment of the initial features with the target function $y(x)$. We argue this delayed generalization arises when (1) the top eigenvectors of the initial neural tangent kernel and the task labels $y(x)$ are misaligned, but (2) the dataset size is large enough so that it is possible for the network to generalize eventually, but not so large that train loss perfectly tracks test loss at all epochs, and (3) the network begins training in the lazy regime so does not learn features immediately. We conclude with evidence that this transition from lazy (linear model) to rich training (feature learning) can control grokking in more general settings, like on MNIST, one-layer Transformers, and student-teacher networks.

stat.ML↗

arXiv2024

Successor-Predecessor Intrinsic Exploration

Exploration is essential in reinforcement learning, particularly in environments where external rewards are sparse. Here we focus on exploration with intrinsic rewards, where the agent transiently augments the external rewards with self-generated intrinsic rewards. Although the study of intrinsic rewards has a long history, existing methods focus on composing the intrinsic reward based on measures of future prospects of states, ignoring the information contained in the retrospective structure of transition sequences. Here we argue that the agent can utilise retrospective information to generate explorative behaviour with structure-awareness, facilitating efficient exploration based on global instead of local information. We propose Successor-Predecessor Intrinsic Exploration (SPIE), an exploration algorithm based on a novel intrinsic reward combining prospective and retrospective information. We show that SPIE yields more efficient and ethologically plausible exploratory behaviour in environments with sparse rewards and bottleneck states than competing methods. We also implement SPIE in deep reinforcement learning agents, and show that the resulting agent achieves stronger empirical performance than existing methods on sparse-reward Atari games.

cs.LG↗

arXiv2023

How should the advent of large language models affect the practice of science?

Large language models (LLMs) are being increasingly incorporated into scientific workflows. However, we have yet to fully grasp the implications of this integration. How should the advent of large language models affect the practice of science? For this opinion piece, we have invited four diverse groups of scientists to reflect on this query, sharing their perspectives and engaging in debate. Schulz et al. make the argument that working with LLMs is not fundamentally different from working with human collaborators, while Bender et al. argue that LLMs are often misused and over-hyped, and that their limitations warrant a focus on more specialized, easily interpretable tools. Marelli et al. emphasize the importance of transparent attribution and responsible use of LLMs. Finally, Botvinick and Gershman advocate that humans should retain responsibility for determining the scientific roadmap. To facilitate the discussion, the four perspectives are complemented with a response from each group. By putting these different perspectives in conversation, we aim to bring attention to important considerations within the academic community regarding the adoption of LLMs and their impact on both current and future scientific practices.

cs.CL↗

arXiv2022

The molecular memory code and synaptic plasticity: a synthesis

The most widely accepted view of memory in the brain holds that synapses are the storage sites of memory, and that memories are formed through associative modification of synapses. This view has been challenged on conceptual and empirical grounds. As an alternative, it has been proposed that molecules within the cell body are the storage sites of memory, and that memories are formed through biochemical operations on these molecules. This paper proposes a synthesis of these two views, grounded in a computational theory of memory. Synapses are conceived as storage sites for the parameters of an approximate posterior probability distribution over latent causes. Intracellular molecules are conceived as storage sites for the parameters of a generative model. The theory stipulates how these two components work together as part of an integrated algorithm for learning and inference.

q-bio.NC↗

arXiv2022

Representation learning with reward prediction errors

The Reward Prediction Error hypothesis proposes that phasic activity in the midbrain dopaminergic system reflects prediction errors needed for learning in reinforcement learning. Besides the well-documented association between dopamine and reward processing, dopamine is implicated in a variety of functions without a clear relationship to reward prediction error. Fluctuations in dopamine levels influence the subjective perception of time, dopamine bursts precede the generation of motor responses, and the dopaminergic system innervates regions of the brain, including hippocampus and areas in prefrontal cortex, whose function is not uniquely tied to reward. In this manuscript, we propose that a common theme linking these functions is representation, and that prediction errors signaled by the dopamine system, in addition to driving associative learning, can also support the acquisition of adaptive state representations. In a series of simulations, we show how this extension can account for the role of dopamine in temporal and spatial representation, motor response, and abstract categorization tasks. By extending the role of dopamine signals to learning state representations, we resolve a critical challenge to the Reward Prediction Error hypothesis of dopamine function.

q-bio.NC↗

arXiv2022

Hybrid Memoised Wake-Sleep: Approximate Inference at the Discrete-Continuous Interface

Modeling complex phenomena typically involves the use of both discrete and continuous variables. Such a setting applies across a wide range of problems, from identifying trends in time-series data to performing effective compositional scene understanding in images. Here, we propose Hybrid Memoised Wake-Sleep (HMWS), an algorithm for effective inference in such hybrid discrete-continuous models. Prior approaches to learning suffer as they need to perform repeated expensive inner-loop discrete inference. We build on a recent approach, Memoised Wake-Sleep (MWS), which alleviates part of the problem by memoising discrete variables, and extend it to allow for a principled and effective way to handle continuous variables by learning a separate recognition model used for importance-sampling based approximate inference and marginalization. We evaluate HMWS in the GP-kernel learning and 3D scene understanding domains, and show that it outperforms current state-of-the-art inference methods.

cs.CV↗

arXiv2021

Human-Level Reinforcement Learning through Theory-Based Modeling, Exploration, and Planning

Reinforcement learning (RL) studies how an agent comes to achieve reward in an environment through interactions over time. Recent advances in machine RL have surpassed human expertise at the world's oldest board games and many classic video games, but they require vast quantities of experience to learn successfully -- none of today's algorithms account for the human ability to learn so many different tasks, so quickly. Here we propose a new approach to this challenge based on a particularly strong form of model-based RL which we call Theory-Based Reinforcement Learning, because it uses human-like intuitive theories -- rich, abstract, causal models of physical objects, intentional agents, and their interactions -- to explore and model an environment, and plan effectively to achieve task goals. We instantiate the approach in a video game playing agent called EMPA (the Exploring, Modeling, and Planning Agent), which performs Bayesian inference to learn probabilistic generative models expressed as programs for a game-engine simulator, and runs internal simulations over these models to support efficient object-based, relational exploration and heuristic planning. EMPA closely matches human learning efficiency on a suite of 90 challenging Atari-style video games, learning new games in just minutes of game play and generalizing robustly to new game situations and new levels. The model also captures fine-grained structure in people's exploration trajectories and learning dynamics. Its design and behavior suggest a way forward for building more general human-like AI systems.

cs.AI↗

arXiv2021

Language-Mediated, Object-Centric Representation Learning

We present Language-mediated, Object-centric Representation Learning (LORL), a paradigm for learning disentangled, object-centric scene representations from vision and language. LORL builds upon recent advances in unsupervised object discovery and segmentation, notably MONet and Slot Attention. While these algorithms learn an object-centric representation just by reconstructing the input image, LORL enables them to further learn to associate the learned representations to concepts, i.e., words for object categories, properties, and spatial relationships, from language input. These object-centric concepts derived from language facilitate the learning of object-centric representations. LORL can be integrated with various unsupervised object discovery algorithms that are language-agnostic. Experiments show that the integration of LORL consistently improves the performance of unsupervised object discovery methods on two datasets via the help of language. We also show that concepts learned by LORL, in conjunction with object discovery methods, aid downstream tasks such as referring expression comprehension.

cs.LG↗

arXiv2019

What does the free energy principle tell us about the brain?

The free energy principle has been proposed as a unifying account of brain function. It is closely related, and in some cases subsumes, earlier unifying ideas such as Bayesian inference, predictive coding, and active learning. This article clarifies these connections, teasing apart distinctive and shared predictions.

q-bio.NC↗

arXiv2019

Analyzing machine-learned representations: A natural language case study

As modern deep networks become more complex, and get closer to human-like capabilities in certain domains, the question arises of how the representations and decision rules they learn compare to the ones in humans. In this work, we study representations of sentences in one such artificial system for natural language processing. We first present a diagnostic test dataset to examine the degree of abstract composable structure represented. Analyzing performance on these diagnostic tests indicates a lack of systematicity in the representations and decision rules, and reveals a set of heuristic strategies. We then investigate the effect of the training distribution on learning these heuristic strategies, and study changes in these representations with various augmentations to the training set. Our results reveal parallels to the analogous representations in people. We find that these systems can learn abstract rules and generalize them to new contexts under certain circumstances -- similar to human zero-shot reasoning. However, we also note some shortcomings in this generalization behavior -- similar to human judgment errors like belief bias. Studying these parallels suggests new ways to understand psychological phenomena in humans as well as informs best strategies for building artificial intelligence with human-like language understanding.

cs.CL↗

arXiv2018

Human-in-the-Loop Interpretability Prior

We often desire our models to be interpretable as well as accurate. Prior work on optimizing models for interpretability has relied on easy-to-quantify proxies for interpretability, such as sparsity or the number of operations required. In this work, we optimize for interpretability by directly including humans in the optimization loop. We develop an algorithm that minimizes the number of user studies to find models that are both predictive and interpretable and demonstrate our approach on several data sets. Our human subjects results show trends towards different proxy notions of interpretability on different datasets, which suggests that different proxies are preferred on different tasks.

stat.ML↗