Máté Lengyel

This paper introduces a predictive map model that explains multi-scale place field activity in hippocampal neurons and validates its predictions with experimental data.

The paper analyzes a recurrent neural network model of CA3 pyramidal neurons to predict and confirm that their activation function has an exponent slightly above 1, supporting stable memory recall.

This study investigates how implicit vs. explicit context processing affects sensorimotor memory expression under uncertainty, finding that implicit processing leads to belief-weighted combination while explicit processing favors categorical selection.

This study shows how neural circuits adjust their representation of task-relevant and irrelevant stimuli based on noise and metabolic cost, aligning with minimal representational strategies observed in primate prefrontal cortex.

The paper introduces an infinite GPFA model to infer dynamically changing latent factors in neural data, showing it captures meaningful dynamics in hippocampal place cells during spatial tasks.

The paper investigates optimal information loading into working memory in prefrontal cortex using mathematical analysis, simulations, and monkey recordings, finding orthogonal inputs lead to dynamic coding.

The paper shows that spatial uncertainty, influenced by environmental geometry, explains navigational behavior and grid-cell properties through a Bayesian model.

This paper introduces a Bayesian nonparametric method to model neural spike train variability, showing improved performance in capturing interspike interval statistics compared to existing approaches.

This paper trains a biologically realistic neural network model, the stabilized supralinear network, on challenging tasks using a dynamics-neutral growth method during training.

The paper extends the manifold Gaussian process latent variable model to handle non-Euclidean neural data with temporal continuity and count-based noise, improving performance on synthetic and real neural recordings.

The paper develops a theory of motor learning based on contextual inference, showing it explains key features of motor memory formation and predicts novel phenomena.

This paper introduces a probabilistic spike count model to capture neural variability across time and covariates, finding that variability is modulated by external factors and explained by latent factors.

This study develops a recurrent neural network model of a visual cortical hypercolumn optimized for sampling-based probabilistic inference, revealing how cortical dynamics arise from efficient computational functions.

The paper develops a method to model dendritic computations in neurons under in vivo-like conditions and finds that a single global non-linearity captures most of the somatic response.

The paper shows that a stochastic Stabilized Supralinear Network model explains stimulus-induced changes in cortical variability and correlations.