Non-negative matrix factorization for receptive field analysis

Using Non-negative matrix factorization instead of SVD Sujay, a labmate of mine came to me with an interesting analysis problem. He’s looking at perisaccadic changes in receptive fields in V4. The saccade-triggered receptive fields shows two activations: an initial one at the pre-saccadic location, and a later one at the remapped location. Basically, the activation … More Non-negative matrix factorization for receptive field analysis

Why second-order methods can be futile in non-convex problems

I’ve been working on fitting a convolutional model of neurons in primary and intermediate visual cortex. A non-convex optimization problem must be solved to estimate the parameters of the model. It has a form similar to: There are some shared weights to further complicate things, but the most salient features is that it’s a 3-layer … More Why second-order methods can be futile in non-convex problems

Topography conquers all

The eye faithfully maps visual space to different positions on the retina. This retinotopy is preserved as the signal is forwarded from retinal ganglion cells to the LGN, then to V1, and onwards. Cells which are physically adjacent on a retinotopic map have receptive fields corresponding to similar positions in space. More generally, properties like … More Topography conquers all

Using an iterated extended Kalman filter to decode place cells

Decoding neuronal activity is a powerful technique to study how information is encoded in a population and how it might be extracted by other brains areas. Hippocampal place cells are a prime example of a system that can be studied fruitfully from a decoding persepective. In a typical place cell decoding experiment, a population of … More Using an iterated extended Kalman filter to decode place cells