Gradient presentacions in ReLU networks as similarity functions, the targent sensitivity matrix

Mute
Current Time 0:00
/
Duration Time 0:00
Loaded: 0%
Progress: 0%
Stream TypeLIVE
Remaining Time -0:00
 

Description of video

Date: 3/30/22
Speaker :Rácz Dániel (SZTAKI)

Feed-forward networks can be interpreted as mappings with linear decision surfaces at the level of the last layer.
e investigate how the tangent space of the network can be exploited to refine the decision in case of ReLU (Rectified Linear Unit) activations.
We show that a simple Riemannian metric parametrized on the parameters of the network forms a similarity function at least as good as the original network and we
suggest a sparse metric to increase the similarity gap.

Downloads