Benjamin Midler

This study shows evolutionarily conditioned neural networks exhibit unique learning dynamics and can be rapidly fine-tuned, suggesting evolution provides an inductive bias for efficient learning.

This paper introduces a framework for multi-view animal pose estimation that improves accuracy and uncertainty estimation with limited labeled data, validated across multiple species.

This paper shows recurrent neural networks can improve occluded object recognition by using occluder information to 'explain-away' occlusions, as demonstrated through computational models and human experiments.

The paper investigates how direct and indirect basal ganglia pathways influence cortical activity during decision-making, finding that they oppositely modulate a subset of neurons encoding decision variables.

The paper shows how two cell types in Drosophila combine head direction and goal signals to produce motor commands for navigation.

The paper demonstrates that a neural network model with non-monotonic plasticity exhibits retrieval induced forgetting (RIF), supporting the non-monotonic plasticity hypothesis.