Deep Curvature Suite
This provides a practical tool for ML researchers to better understand neural network properties, though it is incremental as it builds on existing Lanczos algorithm theory.
The researchers tackled the underutilization of curvature information in neural networks by developing Deep Curvature Suite, a PyTorch-based open-source package for analyzing and visualizing curvature and loss landscapes, demonstrating its superiority over existing approaches on CIFAR datasets.
We present MLRG Deep Curvature suite, a PyTorch-based, open-source package for analysis and visualisation of neural network curvature and loss landscape. Despite of providing rich information into properties of neural network and useful for a various designed tasks, curvature information is still not made sufficient use for various reasons, and our method aims to bridge this gap. We present a primer, including its main practical desiderata and common misconceptions, of \textit{Lanczos algorithm}, the theoretical backbone of our package, and present a series of examples based on synthetic toy examples and realistic modern neural networks tested on CIFAR datasets, and show the superiority of our package against existing competing approaches for the similar purposes.