# Hyperplane Arrangements of Trained ConvNets Are Biased

Source: [Hyperplane Arrangements of Trained ConvNets Are Biased](<https://www.sota2.com/research/paper/68fa6f071e209131cea43655>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T01:11:39.114Z

Authors: Matteo Gamba, Stefan Carlsson, Hossein Azizpour, M{\\aa}rten Bj\\"orkman

Published: 2020-03-17T16:22:17.000Z

## Abstract

We investigate the geometric properties of the functions learned by trained ConvNets in the preactivation space of their convolutional layers, by performing an empirical study of hyperplane arrangements induced by a convolutional layer. We introduce statistics over the weights of a trained network to study local arrangements and relate them to the training dynamics. We observe that trained ConvNets show a significant statistical bias towards regular hyperplane configurations. Furthermore, we find that layers showing biased configurations are critical to validation performance for the architectures considered, trained on CIFAR10, CIFAR100 and ImageNet.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2003.07797>)

[PDF](<https://arxiv.org/pdf/2003.07797.pdf>)