Machine Learning [Please read the instructions below the introduction.] [This seems like it reached a terrible local minimum point, and that it didn't work. I'll try to take a look at it.] [Also, the weights won't update unless cloud lists exist on Scratch, and that won't happen.] > Introduction Sometimes doing complex recognition tasks take a lot of work for people to program. But what if we can program the algorithm to learn by itself? In this *very* complex Scratch program, I coded Artificial Intelligence. More specifically, I made neural networks to make the machine teach itself. Handwriting is very hard to recognize, but I taught the algorithm to learn by itself. > Instructions Draw a digit. Any digit you want from 0 to 9. Press the Space bar to submit. After a while the program will show a list. The values show how much the algorithm believes the number is. For example, if the 5th number in the list is 0.985, then 5 is likely your digit drawn. Please note that the 10th number in the list denotes the 0. In order to make the algorithm improve, lots of data must be entered. I've done lots of data, but I need your help. After the values fully display, press the number you drew on the keyboard. When the green flag stops running, you are done with the program and can press it again for more results. Note that, because this is a Scratch project that likely will never have more than a thousand views, it would never have enough data, and would never be actually that accurate. Enjoy! This took a long time to program.
If you came here from Scratch, this is a school project. If you came from school, this is a Scratch project. #MachineLearning #ArtificialIntelligence #AI