Watch the neural network learn! Turn on turbo mode (shift-click the green flag) for faster training (although it's pretty fast, so you might not see much). As the outputs become closer to the target values (shown inside the output nodes), the cost decreases, so the target cost is zero. A red line represents a negative weight value, green a positive, a thicker line represents a larger weight (in terms of absolute value), thinner a smaller.
Note that this is not a very official project, I just want to demonstrate what I've been working on. The more official projects will probably come later in the form of an image classifier (hand-drawn numbers 0-9) and a relatively user-friendly neural network creator. Thank you to @TheMinerMan44 for the color demonstration idea ( https://scratch.mit.edu/projects/104660889/ ).
