Hi Luke, the thing is that it’s normal behavior for neural networks, since optimization process is basically stochastic. You can help it a bit with fixing a random seed + you have to run training several times and check mean and standard deviation of your loss

Co-founder of consulting firm Neurons Lab and advisor to AI products builders. On Medium, I write about proven strategies for achieving ML technology leadership

Co-founder of consulting firm Neurons Lab and advisor to AI products builders. On Medium, I write about proven strategies for achieving ML technology leadership