Machine Learning

• This repository contains various projects which have been solved with the help of machine learning
• This link has a class that uses Optuna package to optimize the hyperparameters of transformers, MLP, XGB, CNN using the TPE algorithm.

Higgs Boson Challenge

• The class was created to be used for the hyperparameters optimization in the google colab. Google's colab is a free-of-cost Platform as a service cloud model.
• The colab uses the above 4 different ML algorithms for the Higgs selection from the background and the data comes from the HiggsChallenge
  • The colab also shows the inner workings of the ML model by using the Shapley score using the SHAP library. The SHAP method fits a simplified model on the ML model locally. The contribution of an individual variable is the difference between its presence and absence while predicting a label for a class.
• The offline notebook is available at the link

Flowers classification

• To classify flowers from 5 different species CNN is used google colab
• The notebook uses TPE algorithm for the hyperparameter optimization
• The final CNN model is compared on an example with the object detection model from cvlib library

Running the code on your personal computer

• If you want to run it on your personal computer e.g. on Visual Studio or terminal then
• cd /directory/on/your/personal/computer
  • Clone the repository:

    git clone https://github.com/shahidzk1/Machine_learning.git
    

  • cd /directory/on/your/personal/computer/Machine_learning/

    git pull origin main
    pip install -r requirements.txt
    python setup.py install
    

  • cd /directory/on/your/personal/computer/Machine_learning/test/
  • For unit testing run
    • python test_hp_opt_class_code.py
    • The output should end with

      Ran 4 tests in xs (x is time in sec)
      OK
      

  • Run the notebook
  • If you want to play with the code, then create your branch and move to that branch

        git branch branch_name
        git checkout branch_name
    

  • if you find issues then kindly mention them in the issues
  • if you modified the code for the better, then kindly commit the changes with a comment and then make a merge request

       git add file.py
       git commmit -m "this comment is for the changes xyz"
    
    

Azure Cloud

• Running on Azure cloud, in a terminal on notebooks in Azure AI machine learning studio, can lead to warnings about the scikit-learn version, which can be ignored.
• If optuna and optuna-integration packages are not found after requirements installation then simply use the following

        pip install optuna
        pip install optuna-integration