Validation and model tuning

• validation fn (anomaly / nominal) in separate file ( ~ loss.py)
• test parameters / hyperparameter tuning (systematically using tensorboard?)
  • window_size
  • # epochs
  • Arch
    • # neurons, latent_dim
    • # layer?
  • batch_size
  • dev in rest_pos
• create evaluation step within trainer
• run first hyperparamter analysis