Jun 19, 2017 · We propose a new technique, Singular Vector Canonical Correlation Analysis (SVCCA), a tool for quickly comparing two representations in a way that is both ...

We propose a new technique, Singular Vector Canonical Correlation Analysis. (SVCCA), a tool for quickly comparing two representations in a way that is both.

We propose a new technique, Singular Vector Canonical Correlation Analysis (SVCCA), a tool for quickly comparing two representations in a way that is both ...

This work introduces a technique based on the singular vector canonical correlation analysis (SVCCA) for measuring the generality of neural network layers.

Missing: Improvement. | Show results with:Improvement.

SVCCA: Singular Vector Canonical Correlation Analysis for Deep Understanding and Improvement. June 2017. Authors: Maithra Raghu · Maithra Raghu.

With this interpretation of neurons as vectors (and layers as subspaces, spanned by neurons), we introduce SVCCA, Singular Vector Canonical Correlation Analysis ...

Jun 22, 2017 · I've given it a quick first read, and I think this is an all-around strong paper that offers well-founded motivations for FreezeOut-style ...

Duration: 19:28
Posted: Jun 8, 2021

Feb 18, 2019 · Researchers from Google Brain and DeepMind have extended SVCCA in [1] to investigate further insights of deep neural representations.