CNN-Hyperparameter-optimization

Designing, training and optimizing MLP NN

This project is an assignment made in the context of my Master studies

Technologies

Python
Pytorch
Numpy
Matplolib
Pandas
Google colab

Dataset

Fashion-MNIST dataset (source : https://github.com/zalandoresearch/fashion-mnist)

10 clothes' categories
Training set : 55000 images
Validation set : 5000
Test set : 1000

Roadmap

Data preprocessing
Data augmentation
Make the NN flexible to efficiently swipe hyper parameters
Searching for the best Hyper-parameters
- Preprocess Wandb
- Run four times Wandb. At each run, re adjust Hyper-parameters according to last run results
Evaluate the final model