NeuroMapper: In-browser Visualizer for Neural Network Training

TL;DR

NeuroMapper is an in-browser visualization tool that enables real-time monitoring and interpretation of neural network training dynamics, specifically visualizing embedding evolution across epochs for models like ResNet-50.

Contribution

It introduces a scalable, real-time visualization method for neural network training embeddings, adapting AlignedUMAP for spatial coherence across epochs, and is accessible via web browsers.

Findings

Visualizes 40,000 embedded points in real time

Allows exploration of training dynamics of ResNet-50

Open-sourced and browser-compatible tool

Abstract

We present our ongoing work NeuroMapper, an in-browser visualization tool that helps machine learning (ML) developers interpret the evolution of a model during training, providing a new way to monitor the training process and visually discover reasons for suboptimal training. While most existing deep neural networks (DNNs) interpretation tools are designed for already-trained model, NeuroMapper scalably visualizes the evolution of the embeddings of a model's blocks across training epochs, enabling real-time visualization of 40,000 embedded points. To promote the embedding visualizations' spatial coherence across epochs, NeuroMapper adapts AlignedUMAP, a recent nonlinear dimensionality reduction technique to align the embeddings. With NeuroMapper, users can explore the training dynamics of a Resnet-50 model, and adjust the embedding visualizations' parameters in real time. NeuroMapper is open-sourced at https://github.com/poloclub/NeuroMapper and runs in all modern web browsers. A demo of the tool in action is available at: https://poloclub.github.io/NeuroMapper/.