boudenjal-mohcine / atelier1_DL

Atelier 1

PyTorch Regression and Multi-Class Classification Lab

Objective

This lab aims to familiarize with PyTorch library for Regression and Multi-Class Classification tasks using Deep Neural Networks (DNNs)/Multi-Layer Perceptrons (MLPs).

Regression

1. Exploratory Data Analysis (EDA)

• Understand and visualize the dataset to gain insights.

2. Deep Neural Network Architecture

• Implement a DNN architecture using PyTorch for regression.

3. Hyperparameter Tuning

• Use GridSearch from sklearn to find optimal hyperparameters.

4. Visualization

• Plot Loss and Accuracy against Epochs for both training and test data.

5. Regularization

• Apply various regularization techniques and compare with the initial model.

Multi-Class Classification

1. Data Preprocessing

• Clean and standardize/normalize the data.

2. Exploratory Data Analysis (EDA)

• Explore and visualize the dataset.

3. Data Augmentation

• Apply techniques to balance the dataset.

4. Deep Neural Network Architecture

• Design a DNN architecture using PyTorch for multi-class classification.

5. Hyperparameter Tuning

• Utilize GridSearch to find optimal hyperparameters.

6. Visualization

• Plot Loss and Accuracy against Epochs for both training and test data.

7. Metrics Calculation

• Evaluate metrics like accuracy, sensitivity, and f1 score on training and test datasets.

8. Regularization

• Apply various regularization techniques and compare with the initial model.

Datasets

• Regression: NYSE Dataset
• Multi-Class Classification: Machine Predictive Maintenance Dataset