Tahra L. Eissa

Task-Parametrized Dynamics: Representation of Time and Decisions in Recurrent Neural Networks

This study investigates how RNNs represent time and make decisions during delayed tasks, finding multiple dynamical solutions with similar behavior.

Learning efficient representations of environmental priors in working memory

The paper shows that human working memory biases are better explained by experience-dependent learning models than by models without learned biases.

Suboptimal human inference inverts the bias-variance trade-off for decisions with asymmetric evidence

The paper shows that humans use suboptimal inference strategies in rare-event tasks, inverting the bias-variance trade-off compared to optimal Bayesian strategies.