NOTE: This section is autogenerated by readme_ation, manual updates will be overwritten
This project implements a neural network model to classify handwritten digits from the MNIST dataset. It explores various techniques to improve model performance, including hyperparameter tuning and data augmentation.
To gain hands-on experience with neural network implementation, hyperparameter optimization, and strategies for improving model accuracy on a classic machine learning problem.
Python, PyTorch, fastai, pandas, matplotlib, seaborn
The project follows these main steps:
- Data preparation and exploration,
- Implementing a simple neural network model,
- Grid search for hyperparameter optimization,
- Analyzing misclassified examples,
- Attempting to improve the model by augmenting the training data with misclassified examples.
Tuning hyperparameters effectively, handling large datasets, implementing data augmentation techniques, and interpreting results to improve model performance.
Grid search can effectively optimize hyperparameters. Visualizing and analyzing misclassified examples provides insights into model weaknesses. However, simply augmenting training data with misclassified examples does not guarantee improved performance.
The project uses the fastai library and follows some concepts from the fastai course. It also uses a custom package readme_ation for adding setup details to the README.
NOTE: This section is autogenerated by readme_ation, manual updates will be overwritten
This will guide you through the process of setting up a Mamba environment and running the provided Python code to see it in action. It uses the last known working versions of Python and packages used.
Ensure you have Mamba installed on your system. Mamba is a fast, cross-platform package manager.
-
Create a Mamba Environment
Open your terminal and execute the following commands:
mamba create -n [ENVIRONMENT_NAME] python=3.9.19 -y mamba activate [ENVIRONMENT_NAME]
-
Install Necessary Packages
# Install each with mamba and fall back to pip if necessary for pkg in PIL=10.3.0, csv=1.0, fastai, itertools=10.1.0, matplotlib=3.8.4, pandas, pkg_vers=0.0.8, readme_ation=0.1.14, seaborn; do (mamba install $pkg -y || pip install "${pkg//=/==}"); done
-
Run the Script
Ensure you are in your project directory and run:
python [FILE_NAME]
Or click 'run' in your IDE of choice.