A project exploiting KNN & K-means clustering to recognize handwritten numbers (inspired by Yann Lecun's work)
Note: To be able to run the program you need Java 1.10 at least.
To run the program just run the KNN.java file. To change and show the number of predicted handwritten numbers, change the 'TESTS' variable (line 8) to the number you want and change the last two parameters of the call to Helpers.show (line 24) so that the first multiplied by the second equals 'TESTS'. (Example : TESTS = 100, first param = 10, second param = 10 - It will then have a grid with 10 rows and 10 colums showing the results.) To tweak the classification, play with the variable 'K'. Be careful to not overfit or underfit! Have fun :)