# Проведіть експерименти з різними архітектурами нейронних мереж
# * No convolutional layers
# * different normalization types (BatchNorm, LayerNorm, InstanceNorm)
# * Dropout
# * Different epochs [5, 10, 15, 20]
# * Different features

# Homework: 
# 1. Overfit on one batch
# Result: loss: 0.0000, accuracy: 1.0

# 2. Simple train
# Result:
# Test accuracy: 78.95%
# [100] loss: 0.6216, accuracy: 0.7854310274124146