ml-decision-surfaces-lab: A Zero-Code Interactive Experiment Platform for Visualizing Machine Learning Decision Boundaries

This is an open-source tool based on Gradio and scikit-learn, designed to help beginners intuitively understand the decision boundaries and regression surfaces of machine learning algorithms (like decision trees, SVM, logistic regression) without writing any code. It addresses the 'black box' barrier faced by ML learners by providing an interactive visual experience.

Machine learning algorithms' abstract nature often hinders beginners. Questions like how decision trees split data, where SVM's hyperplane lies, or why neural network boundaries are complex are hard to grasp. This project aims to solve this pain point by turning abstract concepts into visual, interactive experiences.

Created by developer dheepatel01, the project uses Gradio, scikit-learn, and Matplotlib. It follows the MIT open-source license and supports Windows, macOS, and Linux. Users can run it locally with pre-compiled packages (no setup needed), making it ideal for teaching, self-learning, and algorithm demos.

1. Decision Boundary Visualization: Supports algorithms like decision trees, SVM, logistic regression. Users can adjust parameters (e.g., SVM's C value, decision tree's max depth) to see real-time changes in boundaries.
2. Regression Surface Exploration: Allows exploring regression tasks (linear, polynomial) and understanding bias-variance tradeoff via parameter adjustments.
3. Dataset Support: Built-in classic datasets plus custom CSV uploads for personalized experiments.

• Classroom Teaching: Teachers can preset parameters for demos, invite student interaction, or export visuals for courseware.
• Self-Learning: Provides a safe environment to focus on algorithm essence without code errors.
• Algorithm Selection: Helps predict the suitability of linear models, the need for kernel tricks, or an algorithm's sensitivity to outliers in real projects.

The tech stack includes Gradio (web interface), scikit-learn (algorithm backend), and Matplotlib (plotting). This design makes it easy for Python developers to extend: add new algorithms, customize visuals, or contribute via open-source community.

Current limitation: Focuses on 2D feature space (hard to visualize high dimensions directly; uses low-dimensional projections for high-dimensional data). Future plans: Support PCA/t-SNE for high-dimensional data, add deep learning model exploration, animate training process, and include more metrics (ROC curves, confusion matrices).

ml-decision-surfaces-lab turns complex ML principles into intuitive visual experiences, lowering the learning barrier for AI talent. Whether you're a teacher, student, or practitioner, it's worth trying. Project link: https://github.com/dheepatel01/ml-decision-surfaces-lab