One pixel attack for fooling deep neural networks

From statwiki
Jump to navigation Jump to search

Presented by

1. Ziheng Chu

2. Minghao Lu

3. Qi Mai

4. Qici Tan

Introduction

Methodology

We first formalize the generation of adversarial images as an constrained optimization problem. Let [math]\displaystyle{ f_t }[/math]

Evaluation and Results

Measurement Metrics and Datasets

Results

Discussion and Future Work