Interactive Neural Networks

This interactive article introduces neural networks in a way that’s intuitive to understand, whether you’re a curious mind new to the topic, or have used machine learning models but never fully grasped what’s going on under the hood.

It will first dive into the reasoning behind neural network training processes. Through a series of thought-provoking questions, it will break down complex concepts into simple analogies, and answer why certain steps are necessary and what purpose they serve.

Then it will open up the black box and show you what is exactly happening inside neural networks. You will get hands-on with interactive components, which operates on a real neural network and allows you to adjust parameters, see real-time results, and deepen your understanding as you explore. Instead of feeling “disconnected” from the learning process of neural networks, I want to give greater control and freedom to explore for every audience.

If you are interested… No previous knowledge of machine learning is needed—just start exploring with an open mind and a bit of curiosity!

Q: How does a single node predict?

Let’s adjust the weights to see how they affect the final result.

×0.39

×-0.11

×0.06

Size: 0.8

Color: -0.9

Label: Samoyed (1)

1

0.80×0.39+-0.90×-0.11 + 0.06 = 0.57

Step

Train

0.20.10.050.02

∆undefined=-

∂L/∂a

×

∂a/∂z

×

∂z/∂undefined

×

lr

1. Click on different data points.

2. The features and labels will be fed into our Neural Network.

3. Compare the prediction and the label (right answer).

Is the prediction good?

4. Now try to hit the Step Button multiple times, has the result improved for current data point?

How about other data points?

React Flow

Press enter or space to select a node.You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.

Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.