DNN Demo - AND Gate

What's Happening in This DNN?

Input: You choose values for A and B (0 or 1).
Hidden Layer: Two neurons use weighted sums of inputs, passed through a ReLU activation function.
Output Layer: Combines hidden outputs into a final value using the sigmoid function (which outputs a probability).
Loss Calculation: Measures how far the prediction is from the expected output (AND logic).
Backpropagation: Computes error gradients and updates all weights and biases to reduce future error.
Training: Each click of “Simulate Epoch” is one learning step. The model stops when it’s confident.

Select input values for A and B:

A: B:

Weights:
Biases:

Output: -

z1 (pre-ReLU): -

z2 (pre-Sigmoid): -

Weights:
Bias:

Output: -

Expected Output: Based on AND logic

Loss: -

Epochs: 0