Carving up the ventral stream with Deep Synthesis

Carving Up the Social World with Generics

Generics are sentences such as “tigers are striped,” “lions have manes,” and “mosquitoes carry West Nile virus.” Generics express generalizations, but unlike quantified statements, generics do not carry information about howmany members of the kind or category have the property. For example, if asked “how many ravens are black?” one could reply “all [or some, or most, etc] ravens are black,” bu...

Unwrapping the ventral stream.

Carving Up Reality

1. The Problem of the Many Think of Mont Blanc, with its rabbits and foothills and its slurries of moistened rock. We can carve up the reality around Mont Blanc in different ways. If we are hunters, we might include rabbits as parts of the mountain; if we are geologists we might include only rock, perhaps together with a certain amount of air in the crevices and tunnels which have been formed b...

Deep Neural Networks Reveal a Gradient in the Complexity of Neural Representations across the Ventral Stream.

Converging evidence suggests that the primate ventral visual pathway encodes increasingly complex stimulus features in downstream areas. We quantitatively show that there indeed exists an explicit gradient for feature complexity in the ventral pathway of the human brain. This was achieved by mapping thousands of stimulus features of increasing complexity across the cortical sheet using a deep n...

The Computational Magic of the Ventral Stream

I argue that the sample complexity of (biological, feedforward) object recognition is mostly due to geometric image transformations and conjecture that a main goal of the ventral stream – V1, V2, V4 and IT – is to learn-and-discount image transformations. In the first part of the paper I describe a class of simple and biologically plausible memorybased modules that learn transformations from un...