Logistic_Regression_W_V2

This web application visualizes a logistic regression model with one feature and its corresponding loss landscape. Users can interact with the model by adjusting weight w using a widget, and observe how changes affect the loss function and help reach the global minima. In addition to that you can check the confusion matrix on training data.

Features

• Plot of logistic Regression Model: Visualizes how the model fits a given dataset with one feature.

• Loss Landscape Plot: Shows the Binary Cross Entropy Loss (BCE Loss) landscape for different values of weight.

• Dataset Generation using widgets : you can select sample number of each class, and add noise to it.

• Threshold Select : you can select the decision boundary between class orange and class purple.

• Interactive Widgets: Adjust weight (w) in real-time and see the effect on the loss function and sigmoid curve.

• Equations Display: Displays the equations used in the plots: sigmoid: Used in the first plot, and the probability of each input(y_hat). Binary Cross Entropy (BCE) Equation: Used in the second plot. Confusion matrix: Third plor.

Installation

To run this application, you need to have Python installed on your machine. Follow the steps below to set up the environment:

1. Clone the repository:

   git clone https://github.com/Hawar-Dzaee/Logistic_Regression_W_V2.git

2. Install the required packages:

   pip install -r requirements.txt

Usage

To run the application, use the following command:

streamlit run main.py

Files

main.py: Contains the code for the web application. LO_W_V2.ipynb : a notebook that follows main.py (to some extend) requirements.txt: Lists the required Python packages. LICENSE : MIT License

Example

Once the application is running, you will see three plots:

1. sigmoid curve fitting the data :

• Shows the S-shaped curve fitting the dataset with one feature(two classes 0 & 1).

2. Loss Landscape Plot:

• Displays Binary Cross Entropy (BCE) landscape.

3. Confusion matrix plot:

• Confusion matrix for orange & purple classes.

Requirements

The application requires the following Python packages:

• Streamlit
• Torch
• Numpy
• Plotly

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Contact For any questions or suggestions, please open an issue or contact [hawardizayee@gmail.com].