International Journal of Scientific Research and Engineering Development

The pedagogical transition from mathematical derivations to practical intuition in Machine Learning (ML) and Deep Learning (DL) presents a substantial barrier for novice practitioners. Traditional educational resources rely heavily on static diagrams, which fail to capture the temporal dynamics of stochastic gradient descent or the evolving topology of decision boundaries during training. To address this disconnect, we architected the Machine Learning Visualizer, a comprehensive Single Page Application (SPA) built on the React.js ecosystem.This platform moves beyond isolated algorithm demonstrations by implementing a client-side execution engine that runs training loops directly in the browser, ensuring zero-latency feedback. The system features interactive modules for Linear Regression, Decision Boundaries (visualizing K-Nearest Neighbors, SVM, and Logistic Regression), and Deep Learning architectures (including MLP for XOR problems and CNNs for digit recognition). A primary contribution of this work is the development of a novel Model Comparison Dashboard, which enables students to execute multiple algorithms simultaneously on identical, user-generated datasets. This paper details the system’s component-based architecture, the optimization of render cycles for real-time visualization, and the utility of side-by-side metric analysis (MSE, Accuracy, Loss) in demonstrating the trade-offs between model complexity and overfitting.

Syed Zeeyan, Jayashree M, Yashas T.M, Zaid Ilyasi, "An Interactive Visualizer for Machine Learning and Deep Learning Algorithms" International Journal of Scientific Research and Engineering Development, V8(6): Page(2518-2526) Nov-Dec 2025. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.