Loss

class ampligraph.latent_features.Loss(eta, hyperparam_dict, verbose=False)

Abstract class for loss function.

Methods

__init__(eta, hyperparam_dict[, verbose])

Initialize Loss.

get_state(param_name)

Get the state value.

_init_hyperparams(hyperparam_dict)

Initializes the hyperparameters needed by the algorithm.

_inputs_check(scores_pos, scores_neg)

Creates any dependencies that need to be checked before performing loss computations

apply(scores_pos, scores_neg)

Interface to external world.

_apply(scores_pos, scores_neg)

Apply the loss function.

__init__(eta, hyperparam_dict, verbose=False)

Initialize Loss.

Parameters
  • eta (int) – number of negatives

  • hyperparam_dict (dict) – dictionary of hyperparams. (Keys are described in the hyperparameters section)

get_state(param_name)

Get the state value.

Parameters

param_name (string) – Name of the state for which one wants to query the value.

Returns

The value of the corresponding state.

Return type

param_value

_init_hyperparams(hyperparam_dict)

Initializes the hyperparameters needed by the algorithm.

Parameters

hyperparam_dict (dictionary) – Consists of key value pairs. The Loss will check the keys to get the corresponding params.

_inputs_check(scores_pos, scores_neg)

Creates any dependencies that need to be checked before performing loss computations

Parameters
  • scores_pos (tf.Tensor) – A tensor of scores assigned to positive statements.

  • scores_neg (tf.Tensor) – A tensor of scores assigned to negative statements.

apply(scores_pos, scores_neg)

Interface to external world. This function does the input checks, preprocesses input and finally applies loss function.

Parameters
  • scores_pos (tf.Tensor) – A tensor of scores assigned to positive statements.

  • scores_neg (tf.Tensor) – A tensor of scores assigned to negative statements.

Returns

loss – The loss value that must be minimized.

Return type

tf.Tensor

_apply(scores_pos, scores_neg)

Apply the loss function. Every inherited class must implement this function. (All the TF code must go in this function.)

Parameters
  • scores_pos (tf.Tensor) – A tensor of scores assigned to positive statements.

  • scores_neg (tf.Tensor) – A tensor of scores assigned to negative statements.

Returns

loss – The loss value that must be minimized.

Return type

tf.Tensor