PGDVisualisation

Overview

Projected Gradient Descent (PGD) attacks are a type of adversarial attack used to test the robustness of machine learning models. This repository contains visualization tools to help researchers explain and understand PGD attacks.

To-Do List

1. Modify PlotCosineSimilarityGradExample

   • Use actual X and Y data from the PCAClass plot.
   • Modify the radius of the ring plotted to map perturbations of an input.

2. Replicate PlotCosineSimilarityGradExample as a Web Interface

   • Create a web interface where green, yellow, and blue points are draggable objects that update the graph in real-time.

Getting Started

To get started with the visualization tools, follow the instructions below:

1. Clone the repository:

   git clone https://github.com/yourusername/PGDVisualisation.git

2. Navigate to the project directory:

   cd PGDVisualisation

3. Install the required dependencies:

   pip install -r requirements.txt