Evaluating the Impact of Network Depth and Bottleneck Capacity on MNIST Denoising

About this project

Deep dive into Denoising Autoencoders using MNIST. I tested 9 models to see how depth and PCA-based bottlenecks affect cleanup. Higher depth made sharper images but higher MSE—a classic perception-distortion tradeoff.
Project Details
  • Category Machine Learning
  • File Type pdf
Download File
Related Work
The Robustness of SVM Kernels to Noise: A Comparative Analysis on MNIST
View Details