Module monk.tf_keras_1.losses.retrieve_loss

Functions

def retrieve_loss(system_dict)

Retrieve loss post state changes

Args

system_dict : dict

System dictionary storing experiment state and set variables

Returns

dict

updated system dict

Expand source code

def retrieve_loss(system_dict):
    '''
    Retrieve loss post state changes

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables

    Returns:
        dict: updated system dict
    '''
    system_dict["local"]["criterion"] = system_dict["hyper-parameters"]["loss"]["name"];
    name = system_dict["local"]["criterion"];

    if(name == "l1"):
        system_dict["local"]["criterion"] = krlo.mean_absolute_error

    elif(name == "l2"):
        system_dict["local"]["criterion"] = krlo.mean_squared_error

    elif(name == "crossentropy"):
        system_dict["local"]["criterion"] = krlo.categorical_crossentropy;

    elif(name == "binarycrossentropy"):
        system_dict["local"]["criterion"] = krlo.binary_crossentropy;

    elif(name == "kldiv"):
        system_dict["local"]["criterion"] = krlo.kullback_leibler_divergence;

    elif(name == "hinge"):
        system_dict["local"]["criterion"] = krlo.hinge;

    elif(name == "squaredhinge"):
        system_dict["local"]["criterion"] = krlo.squared_hinge;

    return system_dict;