# Understanding Neural Networks via Feature Visualization: A survey

## TL;DR

This survey reviews activation maximization techniques in neural networks, discussing their probabilistic foundations and applications in model debugging and interpretability, highlighting recent advances in feature visualization methods.

## Contribution

It provides a comprehensive review of existing activation maximization methods, introduces a probabilistic perspective, and discusses their applications in neural network understanding.

## Key findings

- Summarizes key activation maximization techniques
- Introduces a probabilistic interpretation of AM methods
- Highlights applications in debugging and explaining networks

## Abstract

A neuroscience method to understanding the brain is to find and study the preferred stimuli that highly activate an individual cell or groups of cells. Recent advances in machine learning enable a family of methods to synthesize preferred stimuli that cause a neuron in an artificial or biological brain to fire strongly. Those methods are known as Activation Maximization (AM) or Feature Visualization via Optimization. In this chapter, we (1) review existing AM techniques in the literature; (2) discuss a probabilistic interpretation for AM; and (3) review the applications of AM in debugging and explaining networks.