arXiv subjects

Michael J. Frank

Publications and source records attributed to Michael J. Frank.

6 recordsLinked to original sources

arXiv2026

Efficient Inference in First Passage Time Models

First passage time models describe the time it takes for a random process to exit a region of interest and are widely used across various scientific fields. Fast and accurate numerical methods for computing the likelihood function in these models are essential for efficient statistical inference of model parameters. Specifically, in computational cognitive neuroscience, generalized drift diffusion models (GDDMs) are an important class of first passage time models that describe the latent psychological processes underlying simple decision-making scenarios. GDDMs model the joint distribution over choices and response times as the first hitting time of a one-dimensional stochastic differential equation (SDE) to possibly time-varying upper and lower boundaries. They are widely applied to extract parameters associated with distinct cognitive and neural mechanisms. However, current likelihood computation methods struggle in common application scenarios in which drift rates dynamically vary within trials as a function of exogenous covariates (e.g., brain activity in specific regions or visual fixations). In this work, we propose a fast and flexible algorithm for computing the likelihood function of GDDMs based on a large class of SDEs satisfying the Cherkasov condition. Our method divides each trial into discrete stages, employs fast analytical results to compute stage-wise densities, and integrates these to compute the overall trial-wise likelihood. Numerical examples demonstrate that our method not only yields accurate likelihood evaluations for efficient statistical inference, but also considerably outperforms existing approaches in terms of speed.

stat.ME↗

arXiv2026

Time-Averaged Drift Approximations are Inconsistent for Inference in Drift Diffusion Models

Drift diffusion models (DDMs) have found widespread use in computational neuroscience, cognitive science, mathematical psychology as well as other fields. They model evidence accumulation in simple decision tasks as a stochastic process drifting towards decision barriers. In models where the drift is both time-varying within a trial and variable across trials, the high computational cost for accurate likelihood evaluation has often led to the use of a computationally convenient surrogate for parameter inference, the time-averaged drift approximation (TADA). In each trial, TADA assumes that the time-varying drift rate can be replaced by its temporal average throughout the trial. This approach enables fast parameter inference using analytical likelihood formulas for DDMs with constant drift. In this work, we show that such an estimator is inconsistent: it does not converge to the true drift, posing a risk of biasing scientific conclusions when parameter estimates are obtained by TADA and similar approximations. We provide an elementary proof of this inconsistency in what is perhaps the simplest possible setting: a Brownian motion with piecewise constant drift hitting a one-sided upper boundary. Furthermore, numerical examples based on an attentional DDM (aDDM) show that using TADA leads to systematic misestimation of attentional effects in decision making and can lead to false conclusions in scientific hypothesis testing.

stat.ME↗

arXiv2025

The dynamic interplay between in-context and in-weight learning in humans and neural networks

Human learning embodies a striking duality: sometimes, we appear capable of following logical, compositional rules and benefit from structured curricula (e.g., in formal education), while other times, we rely on an incremental approach or trial-and-error, learning better from curricula that are randomly interleaved. Influential psychological theories explain this seemingly disparate behavioral evidence by positing two qualitatively different learning systems -- one for rapid, rule-based inferences and another for slow, incremental adaptation. It remains unclear how to reconcile such theories with neural networks, which learn via incremental weight updates and are thus a natural model for the latter type of learning, but are not obviously compatible with the former. However, recent evidence suggests that metalearning neural networks and large language models are capable of "in-context learning" (ICL) -- the ability to flexibly grasp the structure of a new task from a few examples. Here, we show that the dynamic interplay between ICL and default in-weight learning (IWL) naturally captures a broad range of learning phenomena observed in humans, reproducing curriculum effects on category-learning and compositional tasks, and recapitulating a tradeoff between flexibility and retention. Our work shows how emergent ICL can equip neural networks with fundamentally different learning properties that can coexist with their native IWL, thus offering a novel perspective on dual-process theories and human cognitive flexibility.

cs.NE↗

arXiv2024

Transformer Mechanisms Mimic Frontostriatal Gating Operations When Trained on Human Working Memory Tasks

Models based on the Transformer neural network architecture have seen success on a wide variety of tasks that appear to require complex "cognitive branching" -- or the ability to maintain pursuit of one goal while accomplishing others. In cognitive neuroscience, success on such tasks is thought to rely on sophisticated frontostriatal mechanisms for selective \textit{gating}, which enable role-addressable updating -- and later readout -- of information to and from distinct "addresses" of memory, in the form of clusters of neurons. However, Transformer models have no such mechanisms intentionally built-in. It is thus an open question how Transformers solve such tasks, and whether the mechanisms that emerge to help them to do so bear any resemblance to the gating mechanisms in the human brain. In this work, we analyze the mechanisms that emerge within a vanilla attention-only Transformer trained on a simple sequence modeling task inspired by a task explicitly designed to study working memory gating in computational cognitive neuroscience. We find that, as a result of training, the self-attention mechanism within the Transformer specializes in a way that mirrors the input and output gating mechanisms which were explicitly incorporated into earlier, more biologically-inspired architectures. These results suggest opportunities for future research on computational similarities between modern AI architectures and models of the human brain.

cs.AI↗

arXiv2022

Reward-Predictive Clustering

Recent advances in reinforcement-learning research have demonstrated impressive results in building algorithms that can out-perform humans in complex tasks. Nevertheless, creating reinforcement-learning systems that can build abstractions of their experience to accelerate learning in new contexts still remains an active area of research. Previous work showed that reward-predictive state abstractions fulfill this goal, but have only be applied to tabular settings. Here, we provide a clustering algorithm that enables the application of such state abstractions to deep learning settings, providing compressed representations of an agent's inputs that preserve the ability to predict sequences of reward. A convergence theorem and simulations show that the resulting reward-predictive deep network maximally compresses the agent's inputs, significantly speeding up learning in high dimensional visual control tasks. Furthermore, we present different generalization experiments and analyze under which conditions a pre-trained reward-predictive representation network can be re-used without re-training to accelerate learning -- a form of systematic out-of-distribution transfer.

cs.LG↗

arXiv2012

A computational model of inhibitory control in frontal cortex and basal ganglia

Planning and executing volitional actions in the face of conflicting habitual responses is a critical aspect of human behavior. At the core of the interplay between these two control systems lies an override mechanism that can suppress the habitual action selection process and allow executive control to take over. Here, we construct a neural circuit model informed by behavioral and electrophysiological data collected on various response inhibition paradigms. This model extends a well established model of action selection in the basal ganglia by including a frontal executive control network which integrates information about sensory input and task rules to facilitate well-informed decision making via the oculomotor system. Our simulations of the antisaccade, Simon and saccade-override task ensue in conflict between a prepotent and controlled response which causes the network to pause action selection via projections to the subthalamic nucleus. Our model reproduces key behavioral and electrophysiological patterns and their sensitivity to lesions and pharmacological manipulations. Finally, we show how this network can be extended to include the inferior frontal cortex to simulate key qualitative patterns of global response inhibition demands as required in the stop-signal task.

q-bio.NC↗