Stephen A. Baccus Lab

J Neurosci 28:6807–6817 · 2008

A retinal circuit that computes object motion

Baccus SA, Ölveczky BP, Manu M, Meister M

Why this matters: Connects a retinal motion computation to a candidate inhibitory circuit.

DOI: 10.1523/JNEUROSCI.4206-07.2008

Nature 436:71–77 · 2005

Dynamic predictive coding by the retina

Hosoya T, Baccus SA, Meister M

Why this matters: Shows how retinal encoding changes with stimulus statistics.

DOI: 10.1038/nature03689

Neuron 111:2742–2755 · 2023

Interpreting the retinal neural code for natural scenes: from computations to neurons

Maheswaranathan N, McIntosh LT, Tanaka H, Grant S, Kastner DB, Melander JB, Nayebi A, Brezovec LE, Wang JH, Ganguli S, Baccus SA

Why this matters: Connects predictive retinal models to interpretable computations.

DOI: 10.1016/j.neuron.2023.06.007

Introductory

Build a center-surround vision model

Computational / literature-data study

Simulate responses to local contrast and global motion; compare model outputs across receptive-field settings.

Background

Retina, Python, Neural coding

Data

Synthetic images and motion sequences generated in a notebook.

Output

Reproducible notebook or evidence table + research poster

Independent educational idea, not offered or supervised by the lab. Use public or synthetic data only; no wet-lab, animal, clinical or human-subject procedures.

Introductory

Compare adaptation and sensitization models

Computational / literature-data study

Generate synthetic spike trains for changing contrast and evaluate two declared response models.

Background

Adaptation, Statistics

Data

Synthetic stimuli and spike trains.

Output

Reproducible notebook or evidence table + research poster

Independent educational idea, not offered or supervised by the lab. Use public or synthetic data only; no wet-lab, animal, clinical or human-subject procedures.

Introductory

When does a predictive model remain interpretable?

Computational / literature-data study

Compare a linear filter with a small neural network on synthetic retinal responses using held-out stimuli.

Background

Interpretable machine learning, Model evaluation

Data

Synthetic retinal responses; record simulation assumptions.

Output

Reproducible notebook or evidence table + research poster

Independent educational idea, not offered or supervised by the lab. Use public or synthetic data only; no wet-lab, animal, clinical or human-subject procedures.