|
Invariant and selective image representations with application to a model of object recognition in visual cortex and deep learning| old_uid | 15045 |
|---|
| title | Invariant and selective image representations with application to a model of object recognition in visual cortex and deep learning |
|---|
| start_date | 2018/05/09 |
|---|
| schedule | 14h30-16h30 |
|---|
| online | no |
|---|
| summary | The primate brain contains a hierarchy of visual areas, dubbed the ventral stream, which rapidly computes object representations that are both specific for object identity and robust against identity-preserving transformations.
In the first part of the talk I conjecture that one of the main computational goals of the ventral stream is to provide a hierarchical representation of new objects/images which is invariant to transformations and selective for recognition and show how such a representation is characterized by small sample complexity and can be learned during development and visual experience.
Further, from a machine learning point of view, I also show how a class of regularized deep convolutional neural networks that mimic the ventral stream architecture can learn such representations. |
|---|
| responsibles | Sarti, Petitot, Nadal |
|---|
| |
|