Add to ORKGSvrha ovog rada je proučavanje algoritama strojnog učenja i metoda prepoznavanja rukopisa koristeći pritom MNIST skup podataka koji sadrži u sebi zapise 60 000 uzoraka brojeva pisanih rukom i skeniranih u 28x28 piksel formatu. Glavni izazov i cilj bilo je prepoznati odgovarajući algoritam koji bi se mogao koristit u rješavanju problema prepoznavanja brojeva.Purpose of this work is to study algorithms of machine learning and seek methods for recognizing handwritten number recognition with the usage of MNIST dataset which has 60 000 patterns of written numbers scanned in 28x28 pixel format. Main challenge and goal was to recognize appropriate algorithm which can be implemented in solving the problem of handwritten recognition