# MLHW1 Report [TOC] ## Digit Classifications ### Wide Network Total Accuracy: 0.933 ![2024-06-16 21 25 07](https://hackmd.io/_uploads/ryvawwnSR.png) ### Deep Network Total Accuracy: 0.948 ![deep_network](https://hackmd.io/_uploads/BkTzDw2HA.png) ## Feature learning **a.** Visaulize the reconstruction results and the first 16 filters ![filters](https://hackmd.io/_uploads/rJ2hGx-tA.png) **b. (bonus)** Apply denoising and dropout mechanism,and visaulize the reconstruction result and the filters. ![reconstruction_noise_free](https://hackmd.io/_uploads/rJ_EXxbFR.png) ![reconstruction_noisy](https://hackmd.io/_uploads/HyCNmebFA.png)