Comparing models of neural representation based on their metric tensors

  • We generalize a previous method (“eigendistortions” - Berardino et al, 2017) to compare neural/neural network models based on their metric tensors. Metric tensors characterize a model’s sensitivity to stimulus perturbations, reflecting both the geometric and stochastic properties of the representation, and providing an explicit prediction of perceptual discriminability.
Read More

Representational dissimilarity metric spaces for stochastic neural networks

  • Existing methods compare deterministic responses (e.g. artificial networks that lack stochastic layers) or averaged responses (e.g., trial-averaged firing rates in biological data). However, these measures of deterministic representational similarity ignore the scale and geometric structure of noise, both of which play important roles in neural computation.
Read More

Quantifying and predicting chromatic thresholds

  • da Fonseca and Samengo (2016) used a neural model combined with Fisher information (a quantification of perceptual thresholds) to predict the pattern of chromatic thresholds measured in human observers. The model assumes linear cone responses paired with Poisson noise. I furthered the analysis, and studied two additional aspects of chromatic perception.
Read More

Delayed normalization model captures disparate nonlinear neural dynamics measured with different techniques in macaque and human V1

  • Voltage-sensitive dye imaging (VSDI) is a powerful method for measuring neural population responses from the cortex of awake, behaving, subjects. We used VSDI to measure the dynamics of neural population responses in macaque V1 to visual stimuli with a wide range of time courses. We built a simple yet flexible delayed normalization model to capture the dynamics of all these measurements.
Read More