<aside> 🤖
“My CPU is a neural-net processor; a learning computer.”
— The Terminator, Terminator 2
</aside>
Neural Net Fun is an interactive lab I made for myself while working through Andrej Karpathy’s introduction to neural networks.
Here’s a video introduction to the tool, though it doesn’t go into as much depth as this tutorial:
First, start the lab here:
The lab starts with this screen. Let’s dive into what it all means.
The goal of this lab’s neural net is to take any x, y coordinate on the 2D plane and correctly classify it as red or blue.
Notice the four big dots. These are your training data, from which the neural net learns. Each big dot is called a training example.
The point mesh—that is, the small dots—are the neural net’s predictions. If the small dots around a big dot are the same color, that’s good—it means that the neural net is predicting with a high level of accuracy.
Accuracy is shown as a number at the bottom of the screen. The screenshot above says Acc: 100%, which means that the neural net’s predictions for each example’s coordinates match the color of the example.
You can modify the training data by holding down your mouse button (or finger, if on a touchscreen). The square near the bottom-left of the screen indicates your current “brush”: