Without using any ML-dedicated python library, I have tried build and trained my very own logistic regression model with L2 regularisation.
I have implemented the following key features to enhance the model performances:
- Calculating gradient from the objective function
- Gradient descent
- Fast gradient descent
- Backtracking
- Cross-Validation for optimal Lambda value
List of helping libraries used:
- pandas
- numpy
- sklearn.model_selection import train_test_split
- sklearn.preprocessing
- matplotlib.pyplot
- sklearn.metrics import accuracy_score