Holly Kular

hkular [at] ucsd [dot] edu

PhD candidate in the Perception and Cognition Lab at University of California San Diego advised by Dr. John Serences.

My research focus is perception and decision making. I apply statistical modeling to human behavioral data and use machine learning methods to train and evaluate recurrent neural network models. I am interested in how populations of neurons encode and manipulate visual information for flexible decision making and behavior.

In my work utilizing RNNs to test neural theories, I explore the contents of learned representations and task dynamics.

news

May 20, 2025	I presented my work Neural networks rely on top-down representations under conditions of uncertainty at the Vision Sciences Society Conference 2025
Jul 17, 2024	Completed the Cold Spring Harbor Laboratories Computational Neuroscience: vision course
Jun 18, 2024	I presented my work Working memory errors from distraction are not influenced by sensory noise at encoding at the Vision Sciences Society Conference 2024
Jun 12, 2024	Successfully proposed my dissertation and advanced to candidacy.
Aug 10, 2023	Completed Steve Luck’s ERP bootcamp. Focusing on EEG ERP experiment design, data collection, and analysis.

selected publications

JOV

Neural networks rely on top-down representations under conditions of uncertainty

In Journal of Vision, 2025

Abs HTML

Statistical regularities in natural visual environments can ve used to form expectations that guide behavior. Hierarchical neural networks utilize expectations in noisy enviornments.
In Prep

Memory recall errors reflect interacting sensory and mnemonic representations

Kular, and Serences

In Prep, 2025
JOV

Working memory errors from distraction are not influenced by sensory noise at encoding

Kular, and Serences

In Journal of Vision, 2024

Abs HTML

It is unclear whether sensory uncertainty during encoding increases susceptibility to interference or whether memories are relatively stable once information is encoded. These data provide evidence that while working memory representations are prone to adaptive distortions during delay periods, these distortions are not modulated by sensory noise during encoding.