| Project | Author | Published |
|---|---|---|
Perceptron Simulator |
Priyangsu |
05 Mar 2025 |
The Perceptron Simulator is an interactive web-based tool designed to visualize and understand the fundamental concepts of a single-layer perceptron. It allows users to manipulate inputs and biases, observing their effects on the output, which is represented by a simulated potentiometer.
- Interactive 4x4 grid representing the input layer.
- Adjustable bias values for each input cell.
- Visual representation of the perceptron's output using a potentiometer.
- Automated training buttons to adjust biases for positive or negative outputs.
- Clear visual feedback through cell and bias highlighting.
The simulator demonstrates the core functionality of a single-layer perceptron:
- Inputs (Cells): The grid cells represent binary inputs (1 or 0).
- Weights (Biases): Adjustable biases act as weights, determining input influence.
- Weighted Sum: The simulator calculates the weighted sum of inputs based on cell states and biases.
- Activation Function (Simplified): The potentiometer's needle position acts as a simplified activation function.
- Output: The potentiometer displays the perceptron's output.
- Training: Training buttons simulate basic weight adjustment.
This simulation helps visualize:
- Linear Separability: How input combinations affect output.
- Weight Adjustment: How biases alter the output.
- Simplified Model: Understanding the basics of a perceptron.
While this simulator uses a 4x4 grid, datasets like MNIST require much larger input layers (784 inputs). Single-layer perceptrons have limited use cases with complex datasets like MNIST due to their limitation to linearly separable problems.
You can try the Perceptron Simulator live at: View Live
Developed by @priyangsubanerjee, this project aims to provide an educational tool for understanding perceptrons.
For contributions, discussions, or concerns, feel free to open an issue or contribute to the repository.