LGFeb 5, 2021

Rethinking Neural Networks With Benford's Law

arXiv:2102.03313v44 citations
AI Analysis

This work proposes a novel early stopping mechanism for neural networks, which could benefit practitioners by reducing the need for a validation set and potentially improving generalization.

The authors propose a new metric, Model Enthalpy (MLH), based on Benford's Law, which measures the closeness of neural network parameters to this law. They demonstrate that MLH is a strong predictor of validation accuracy and can be used for early stopping, potentially outperforming traditional early stopping with a validation set.

Benford's Law (BL) or the Significant Digit Law defines the probability distribution of the first digit of numerical values in a data sample. This Law is observed in many naturally occurring datasets. It can be seen as a measure of naturalness of a given distribution and finds its application in areas like anomaly and fraud detection. In this work, we address the following question: Is the distribution of the Neural Network parameters related to the network's generalization capability? To that end, we first define a metric, MLH (Model Enthalpy), that measures the closeness of a set of numbers to Benford's Law and we show empirically that it is a strong predictor of Validation Accuracy. Second, we use MLH as an alternative to Validation Accuracy for Early Stopping, removing the need for a Validation set. We provide experimental evidence that even if the optimal size of the validation set is known before-hand, the peak test accuracy attained is lower than not using a validation set at all. Finally, we investigate the connection of BL to Free Energy Principle and First Law of Thermodynamics, showing that MLH is a component of the internal energy of the learning system and optimization as an analogy to minimizing the total energy to attain equilibrium.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes