recommenders/models/vae/multinomial_vae.py [16:182]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
class LossHistory(Callback):
    """This class is used for saving the validation loss and the training loss per epoch."""

    def on_train_begin(self, logs={}):
        """Initialise the lists where the loss of training and validation will be saved."""
        self.losses = []
        self.val_losses = []

    def on_epoch_end(self, epoch, logs={}):
        """Save the loss of training and validation set at the end of each epoch."""
        self.losses.append(logs.get("loss"))
        self.val_losses.append(logs.get("val_loss"))


class Metrics(Callback):
    """Callback function used to calculate the NDCG@k metric of validation set at the end of each epoch.
    Weights of the model with the highest NDCG@k value is saved."""

    def __init__(self, model, val_tr, val_te, mapper, k, save_path=None):

        """Initialize the class parameters.

        Args:
            model: trained model for validation.
            val_tr (numpy.ndarray, float): the click matrix for the validation set training part.
            val_te (numpy.ndarray, float): the click matrix for the validation set testing part.
            mapper (AffinityMatrix): the mapper for converting click matrix to dataframe.
            k (int): number of top k items per user (optional).
            save_path (str): Default path to save weights.
        """
        # Model
        self.model = model

        # Initial value of NDCG
        self.best_ndcg = 0.0

        # Validation data: training and testing parts
        self.val_tr = val_tr
        self.val_te = val_te

        # Mapper for converting from sparse matrix to dataframe
        self.mapper = mapper

        # Top k items to recommend
        self.k = k

        # Options to save the weights of the model for future use
        self.save_path = save_path

    def on_train_begin(self, logs={}):
        """Initialise the list for validation NDCG@k."""
        self._data = []

    def recommend_k_items(self, x, k, remove_seen=True):
        """Returns the top-k items ordered by a relevancy score.
        Obtained probabilities are used as recommendation score.

        Args:
            x (numpy.ndarray, int32): input click matrix.
            k (scalar, int32): the number of items to recommend.

        Returns:
            numpy.ndarray: A sparse matrix containing the top_k elements ordered by their score.

        """
        # obtain scores
        score = self.model.predict(x)

        if remove_seen:
            # if true, it removes items from the train set by setting them to zero
            seen_mask = np.not_equal(x, 0)
            score[seen_mask] = 0

        # get the top k items
        top_items = np.argpartition(-score, range(k), axis=1)[:, :k]

        # get a copy of the score matrix
        score_c = score.copy()

        # set to zero the k elements
        score_c[np.arange(score_c.shape[0])[:, None], top_items] = 0

        # set to zeros all elements other then the k
        top_scores = score - score_c

        return top_scores

    def on_epoch_end(self, batch, logs={}):
        """At the end of each epoch calculate NDCG@k of the validation set.

        If the model performance is improved, the model weights are saved.
        Update the list of validation NDCG@k by adding obtained value

        """
        # recommend top k items based on training part of validation set
        top_k = self.recommend_k_items(x=self.val_tr, k=self.k, remove_seen=True)

        # convert recommendations from sparse matrix to dataframe
        top_k_df = self.mapper.map_back_sparse(top_k, kind="prediction")
        test_df = self.mapper.map_back_sparse(self.val_te, kind="ratings")

        # calculate NDCG@k
        NDCG = ndcg_at_k(test_df, top_k_df, col_prediction="prediction", k=self.k)

        # check if there is an improvement in NDCG, if so, update the weights of the saved model
        if NDCG > self.best_ndcg:
            self.best_ndcg = NDCG

            # save the weights of the optimal model
            if self.save_path is not None:
                self.model.save(self.save_path)

        self._data.append(NDCG)

    def get_data(self):
        """Returns a list of the NDCG@k of the validation set metrics calculated
        at the end of each epoch."""
        return self._data


class AnnealingCallback(Callback):
    """This class is used for updating the value of β during the annealing process.
    When β reaches the value of anneal_cap, it stops increasing."""

    def __init__(self, beta, anneal_cap, total_anneal_steps):

        """Constructor

        Args:
            beta (float): current value of beta.
            anneal_cap (float): maximum value that beta can reach.
            total_anneal_steps (int): total number of annealing steps.
        """
        # maximum value that beta can take
        self.anneal_cap = anneal_cap

        # initial value of beta
        self.beta = beta

        # update_count used for calculating the updated value of beta
        self.update_count = 0

        # total annealing steps
        self.total_anneal_steps = total_anneal_steps

    def on_train_begin(self, logs={}):
        """Initialise a list in which the beta value will be saved at the end of each epoch."""
        self._beta = []

    def on_batch_end(self, epoch, logs={}):
        """At the end of each batch the beta should is updated until it reaches the values of anneal cap."""
        self.update_count = self.update_count + 1

        new_beta = min(
            1.0 * self.update_count / self.total_anneal_steps, self.anneal_cap
        )

        K.set_value(self.beta, new_beta)

    def on_epoch_end(self, epoch, logs={}):
        """At the end of each epoch save the value of beta in _beta list."""
        tmp = K.eval(self.beta)
        self._beta.append(tmp)

    def get_data(self):
        """Returns a list of the beta values per epoch."""
        return self._beta
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



recommenders/models/vae/standard_vae.py [17:182]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
class LossHistory(Callback):
    """This class is used for saving the validation loss and the training loss per epoch."""

    def on_train_begin(self, logs={}):
        """Initialise the lists where the loss of training and validation will be saved."""
        self.losses = []
        self.val_losses = []

    def on_epoch_end(self, epoch, logs={}):
        """Save the loss of training and validation set at the end of each epoch."""
        self.losses.append(logs.get("loss"))
        self.val_losses.append(logs.get("val_loss"))


class Metrics(Callback):
    """Callback function used to calculate the NDCG@k metric of validation set at the end of each epoch.
    Weights of the model with the highest NDCG@k value is saved."""

    def __init__(self, model, val_tr, val_te, mapper, k, save_path=None):

        """Initialize the class parameters.

        Args:
            model: trained model for validation.
            val_tr (numpy.ndarray, float): the click matrix for the validation set training part.
            val_te (numpy.ndarray, float): the click matrix for the validation set testing part.
            mapper (AffinityMatrix): the mapper for converting click matrix to dataframe.
            k (int): number of top k items per user (optional).
            save_path (str): Default path to save weights.
        """
        # Model
        self.model = model

        # Initial value of NDCG
        self.best_ndcg = 0.0

        # Validation data: training and testing parts
        self.val_tr = val_tr
        self.val_te = val_te

        # Mapper for converting from sparse matrix to dataframe
        self.mapper = mapper

        # Top k items to recommend
        self.k = k

        # Options to save the weights of the model for future use
        self.save_path = save_path

    def on_train_begin(self, logs={}):
        """Initialise the list for validation NDCG@k."""
        self._data = []

    def recommend_k_items(self, x, k, remove_seen=True):
        """Returns the top-k items ordered by a relevancy score.
        Obtained probabilities are used as recommendation score.

        Args:
            x (numpy.ndarray, int32): input click matrix.
            k (scalar, int32): the number of items to recommend.

        Returns:
            numpy.ndarray: A sparse matrix containing the top_k elements ordered by their score.

        """
        # obtain scores
        score = self.model.predict(x)

        if remove_seen:
            # if true, it removes items from the train set by setting them to zero
            seen_mask = np.not_equal(x, 0)
            score[seen_mask] = 0

        # get the top k items
        top_items = np.argpartition(-score, range(k), axis=1)[:, :k]

        # get a copy of the score matrix
        score_c = score.copy()

        # set to zero the k elements
        score_c[np.arange(score_c.shape[0])[:, None], top_items] = 0

        # set to zeros all elements other then the k
        top_scores = score - score_c

        return top_scores

    def on_epoch_end(self, batch, logs={}):
        """At the end of each epoch calculate NDCG@k of the validation set.
        If the model performance is improved, the model weights are saved.
        Update the list of validation NDCG@k by adding obtained value.
        """
        # recommend top k items based on training part of validation set
        top_k = self.recommend_k_items(x=self.val_tr, k=self.k, remove_seen=True)

        # convert recommendations from sparse matrix to dataframe
        top_k_df = self.mapper.map_back_sparse(top_k, kind="prediction")
        test_df = self.mapper.map_back_sparse(self.val_te, kind="ratings")

        # calculate NDCG@k
        NDCG = ndcg_at_k(test_df, top_k_df, col_prediction="prediction", k=self.k)

        # check if there is an improvement in NDCG, if so, update the weights of the saved model
        if NDCG > self.best_ndcg:
            self.best_ndcg = NDCG

            # save the weights of the optimal model
            if self.save_path is not None:
                self.model.save(self.save_path)

        self._data.append(NDCG)

    def get_data(self):
        """Returns a list of the NDCG@k of the validation set metrics calculated
        at the end of each epoch."""
        return self._data


class AnnealingCallback(Callback):
    """This class is used for updating the value of β during the annealing process.
    When β reaches the value of anneal_cap, it stops increasing.
    """

    def __init__(self, beta, anneal_cap, total_anneal_steps):

        """Constructor

        Args:
            beta (float): current value of beta.
            anneal_cap (float): maximum value that beta can reach.
            total_anneal_steps (int): total number of annealing steps.
        """
        # maximum value that beta can take
        self.anneal_cap = anneal_cap

        # initial value of beta
        self.beta = beta

        # update_count used for calculating the updated value of beta
        self.update_count = 0

        # total annealing steps
        self.total_anneal_steps = total_anneal_steps

    def on_train_begin(self, logs={}):
        """Initialise a list in which the beta value will be saved at the end of each epoch."""
        self._beta = []

    def on_batch_end(self, epoch, logs={}):
        """At the end of each batch the beta should is updated until it reaches the values of anneal cap."""
        self.update_count = self.update_count + 1

        new_beta = min(
            1.0 * self.update_count / self.total_anneal_steps, self.anneal_cap
        )

        K.set_value(self.beta, new_beta)

    def on_epoch_end(self, epoch, logs={}):
        """At the end of each epoch save the value of beta in _beta list."""
        tmp = K.eval(self.beta)
        self._beta.append(tmp)

    def get_data(self):
        """Returns a list of the beta values per epoch."""
        return self._beta
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -