On the limits of neural network explainability via descrambling

TL;DR

This paper characterizes the mathematical solutions for neural network descrambling, revealing how principal components and eigendecompositions can explain hidden layer transformations and improve interpretability.

Contribution

It provides a formal characterization of neural network descrambling solutions using the Brockett function, connecting explainability to eigendecomposition and principal components.

Findings

Descramblers include Fourier basis modes and semantic features.

Eigendecomposition reveals underlying transformations of hidden layers.

SVD relates closely to neural network explainability.

Abstract

We characterize the exact solutions to neural network descrambling--a mathematical model for explaining the fully connected layers of trained neural networks (NNs). By reformulating the problem to the minimization of the Brockett function arising in graph matching and complexity theory we show that the principal components of the hidden layer preactivations can be characterized as the optimal explainers or descramblers for the layer weights, leading to descrambled weight matrices. We show that in typical deep learning contexts these descramblers take diverse and interesting forms including (1) matching largest principal components with the lowest frequency modes of the Fourier basis for isotropic hidden data, (2) discovering the semantic development in two-layer linear NNs for signal recovery problems, and (3) explaining CNNs by optimally permuting the neurons. Our numerical experiments indicate that the eigendecompositions of the hidden layer data--now understood as the descramblers--can also reveal the layer's underlying transformation. These results illustrate that the SVD is more directly related to the explainability of NNs than previously thought and offers a promising avenue for discovering interpretable motifs for the hidden action of NNs, especially in contexts of operator learning or physics-informed NNs, where the input/output data has limited human readability.