Virtual Reality-Based Educational System for Teaching the Single-Layer Perceptron

This project introduces an innovative Virtual Reality (VR) application designed to teach the core concepts of a Single-Layer Perceptron (SLP)—a foundational model in neural networks. The educational system uses immersive, embodied interactions to translate abstract machine learning concepts into intuitive and practical experiences.

Objectives

• Bridge the gap between abstract neural network theories and practical, intuitive learning experiences.
• Leverage embodied interaction and immersive environments to improve student engagement and comprehension.
• Evaluate the educational effectiveness and usability of the system through empirical user studies.

Foundational Research & User Feedback

In the first phase of this project, we developed and evaluated a VR-based educational system that teaches SLP concepts. Through gamified, step-by-step interaction in a factory-inspired environment, users learned how weights, inputs, and activation functions work together in neural computation. A user study with eight participants provided valuable feedback, highlighting both strengths (e.g., interactivity, engagement) and areas for improvement (e.g., navigation and instruction clarity).

Refined VR Learning Environment

This final version was designed to address the limitations of the initial prototype by enhancing navigation, interaction clarity, and instructional integration. The environment supports a seamless and engaging learning experience for understanding SLP.

Training Room: Students first enter a guided training environment where they learn how to interact with objects using natural hand gestures (via hand tracking). This onboarding phase reduces confusion and ensures smooth transition into learning tasks.

Main Room: In a scientifically styled lab environment, learners engage in interactive tasks that simulate how single-layer perceptrons classify inputs and adjust weights. Realistic visuals and immediate feedback reinforce neural computation principles.

Base Model: Interactive Jupyter Notebook

A Jupyter Notebook-based implementation of the Single-Layer Perceptron was developed. It serves as a non-immersive baseline for comparison against the VR system. The notebook guides learners from theory to practice, allowing them to define model parameters, train the perceptron, visualize decision boundaries, and analyze classification outcomes with interactive widgets for real-time tuning. The Jupyter version acts as a control condition for evaluating the pedagogical value added by the VR system. Side-by-side comparisons assess differences in learning outcomes, engagement, and usability.

Downloads

Publications