In this work I demonstrate how using Batch Normalization, learning rate adjusting and choosing the right optimization method can significantly improve the performance of a model. This work was done during completing the course "Deep Learning" by Simon Kozlov (https://dlcourse.ai/).