base_models

Base Models

This directory contains implementations of the Base-MLP, ResNet-18 and Logistic Regression base models used in evaluation in the paper. Model files containing the trained parameters for each base model used in the paper's evaluation are under the base_model_trained_files directory.

Adding a new base model

If you'd like to add a new base model for evaluation, you need to make to perform the following steps:

1. Implement the base model in PyTorch. Your base model must inherit from torch.nn.Module and must implement the __init__ and forward methods. The forward method must take just one parameter, a batch of samples over which a forward pass is performed.

2. Train your base model and save the base model's state dictionary to a file. This may be done using torch.save(my_base_model.state_dict(), "my_file.t7")

3. Create a new configuration file that specifically changes the following parameters:

   1. BaseModel: Specify the classpath of your base model in the "class" field and any arguments required for the __init__ function in the "args" field.
   2. base_model_file: Specify the path to the state dictionary saved in step (2).
   3. base_model_input_size: Specify the input dimensions expected of inputs to the forward method of your base model.

   There are more details on other configuration parameters in the conf directory.