02620-Metagenomics

Spring 2021 Machine Learning for Scientists group project

Executing the code

Using implemented models

Jupyter Notebook example_execution_rqn_210513.ipynb contains a sample run showing how to execute the sampling process and training for implemented models.

Grid Search

Each Python module in packages.gridsearch contains the code for executing a grid search over a single model type.

Plotting Results

Jupyter Notebooks plot_results_rqn_210409.ipynb and results_plot_psk_042521.ipynb contain the code for generating plots from the grid search results.

Project Structure

Code for this project is subdivided into Python packages and Jupyter notebooks. Notebooks were used for testing and plotting results. Python packages were developed for sampling and encoding the data, implementing required models from scratch, and testing model performance via grid search.

packages

• generative_model: Implements Naïve Bayes
• linear_model: Implements Multiclass Logistic Regression
• metagenomics: Implements fragment sampling and encoding
• gridsearch: Implements grid search functionality to test each model over a range of hyperparameters.

notebooks

• example_execution_rqn_210513.ipynb: Example run showing how to execute the sampling process and training for implemented models.
• data_exploratory_rqn_210321.ipynb: Explores original dataset and builds sequence datasets used by this project.
• dev_logistic_regression_rqn_210327.ipynb: Testing of logistic regression implementation
• dev_sampling_encoding_rqn_210325.ipynb: Testing of sampling and encoding
• plot_results_rqn_210409.ipynb: Plots results of grid search
• practice_psk_040421.ipynb: Testing of sampling and encoding
• psk_testing_naivebayes_041821.ipynb: Testing of Naïve Bayes implementation
• results_plot_psk_042521.ipynb: Plot results for grid search for SVM and Naïve Bayes