Hello! This project is a simple AI that learns to recognize handwritten digits using a perceptron (a basic type of machine learning). It works by adjusting “weights” when it makes mistakes, so it slowly gets better at telling numbers apart. PERFORMANCE: With my current setup, the highest accuracy I achieved is about 90%. WEAKNESS: - Works best when the digit’s size is similar to what it learned. - Certain numbers, like 9 and 8, are very hard to tell apart because they use very similar pixels. - To improve this further, we’d need a more advanced machine learning approach. TIP: Press the space bar to turn on Auto-Learning Mode. This will train it quickly using a built-in dataset. For even faster training, turn on Turbo Mode in Scratch (and remember to adjust the time slider!)
Disclaimer: This is a learning project! I’m still newbie to AI, so it may not be perfect, but I made it to understand how a simple perceptron works.
