AMLS

This repository contains the code for the computer vision final project of the ELEC0134 Applied Machine Learning Systems module at UCL.

How to compile

In the AMLS_19-20_Raphael_Angelo_Floresca_SN16011494 folder, compile main.py. The following command line arguments can be specified, otherwise it will run with the following default settings

• --schedule_type: specifies the type of learning rate schedule to run. Specify the specific learning rate schedules for the models in sequential order in the following format (e.g. one_cycle,one_cycle,one_cycle,one_cycle). Accepts none,step,linear,poly and one_cycle. Default: one_cycle,one_cycle,one_cycle,one_cycle
• --epochs: specifies the number of training epochs. Specify the specific epochs for the models in sequential order in the following format (e.g. 10,10,10,10). Default: 10,10,10,10
• --learning_rates: specifies the learning rates. Specify the specific learning rate for the models in sequential order in the following format (e.g. 0.1,0.2,0.1,0.01). Default: 0.03,0.03,0.03,0.03
• --find_lr: specifies whether the learning rate finder should be used. Default: False
• --random_state: specifies a random seed when creating the training, validation and test sets. Default: None
• --model_type: specifies the models used for each task. Specify the specific models for the tasks in sequential order in the following format (e.g. mlp,mlp,mlp,mlp). Accepts mlp,cnn and xception Default: xception,xception,xception,xception.

Output

Check the output folder to find plots produced during training and testing.