Visualization of Neural Networks Training in Real-Time: A Web Based Example
Dane Milišić, Дину Драган, Dušan B. GAJIĆ, Veljko Petrović · 2025
Real-time visualization of neural network training in web applications presents distinct challenges. Existing desktop solutions burden the user's hardware, requiring powerful machines and local model hosting. In contrast, web applications typically lack real-time visualization, showing only static network structures. Proposed web solution overcomes these hurdles by remotely fetching neural network data, easing local resource use, and enhancing accessibility for all user levels. This is especially beneficial for casual users and serves as an educational tool. Solution's effectiveness has been evaluated by analyzing visualization time efficiency and conducting user tests with diverse user profiles. Despite utilizing React.JS for high-quality data presentation, challenges remain in handling large and visualizing datasets. This paper discusses the framework and initial results, proposing further enhancements for optimizing real-time visualization tools in neural network training.