name some hyperparameters in neural network?
learning rate, no of layers, no of hidden units, batch size…
2 ways to do hyperparamters tuning?
try random values (dont use grid if have many hyperparams)
coarse to find: try a few to see which ranges work best then zoom in to thay range
differences between parameters and hyperparameters?
params: estimated from data automatically
hyperparams: manually set, used in processes to help estimate params