Explore perceptron-style decision boundaries, layered networks, and activation flows. A conceptual bridge toward modern neural architectures.
Perceptron Visualizer
Click to add training points:
Left click = Class +1 (Blue)
Right click = Class -1 (Red)
Press Train Perceptron to compute a separating decision boundary.
Neural Network Forward Pass Simulator
This mini deep learning demo uses a tiny feed-forward network:
3 → 4 → 1 (ReLU in hidden layer).
Enter 3 numbers below and observe the activations.
—
—
CNN Filter Visualizer
This demo uses a small 20×20 grayscale “image” drawn into a canvas. Choose a 3×3 kernel (filter) and apply it to see how convolution changes the image. This mimics what happens inside a CNN layer.
Original
Filtered
Backpropagation Explorer
This tool illustrates backpropagation on a minimal model:
Forward: ŷ = w · x
Loss: L = ½ (ŷ − y)²
Enter values for input x, target y, and parameter w, then compute the gradient ∂L/∂w.
—