Module reclab.recommenders.cfnade.cfnade_lib.utils

Util functions for class Cfnade

""" Util functions for class Cfnade"""
from itertools import islice
import numpy as np
import keras
from keras import backend as K
from keras.callbacks import Callback

class DataSet(Callback):
    """
    A datagenerator the feeds data in batches.

    ratings_df: rating matrix, num_iters * num_users, entry is input rating rounded to integer
    batch_size: int, batch size, default is 64
    num_users: int, number of users
    num_items: int, number of items
    mode: int, 0 for train, 1 for eval, 2 for test
    """
    def __init__(self,ratings_df,
        num_users,
        num_items,
        batch_size,
        rating_bucket,
        mode):

        self.num_users = num_users
        self.num_items = num_items
        self.batch_size = batch_size
        self.ratings_df = ratings_df
        self.rating_bucket = rating_bucket
        self.mode = mode

    def generate(self, eval=False):
        """
        a generator function yields ratings_df for each batch

        """
        line_pointer = 0
        while True:
            next_n_data_lines = list(islice(self.ratings_df, line_pointer, line_pointer+self.batch_size))
            if not next_n_data_lines:
                if self.mode == 0 and eval==False:
                    line_pointer = 0
                    next_n_data_lines = list(islice(self.ratings_df, line_pointer, line_pointer+self.batch_size))
                else:
                    break
            input_ranking_vectors = np.zeros((self.batch_size, self.num_users, self.rating_bucket), dtype='int8')
            output_ranking_vectors = np.zeros((self.batch_size, self.num_users, self.rating_bucket), dtype='int8')
            input_mask_vectors = np.zeros((self.batch_size, self.num_users), dtype='int8')
            output_mask_vectors = np.zeros((self.batch_size, self.num_users), dtype='int8')
            for i, line in enumerate(next_n_data_lines):
                user_ids = np.nonzero(line)[0]
                ratings_line = line[line != 0]

                if self.mode == 0 and len(user_ids) != 0:
                    # a random ordered list 0 to len(user_ids)-1
    
                    ordering = np.random.permutation(np.arange(len(user_ids)))
                    random_num = np.random.randint(0, len(ordering))
                    flag_in = (ordering < random_num)
                    flag_out = (ordering >= random_num)
                  
                    input_mask_vectors[i][user_ids] = flag_in
                    output_mask_vectors[i][user_ids] = flag_out

                    for j, (user_id, value) in enumerate(zip(user_ids, ratings_line)):
                        if flag_in[j]:
                            input_ranking_vectors[i, user_id, (value-1)] = 1
                        else:
                            output_ranking_vectors[i, user_id, (value-1)] = 1
                if self.mode == 2:
                    for j, (user_id, value) in enumerate(zip(user_ids, ratings_line)):
                        input_ranking_vectors[i, user_id, (value-1)] = 1                  

            inputs = {
                'input_ratings': input_ranking_vectors,
                'output_ratings': output_ranking_vectors,
                'input_masks': input_mask_vectors,
                'output_masks': output_mask_vectors}

            outputs = {'nade_loss': np.zeros([self.batch_size])}
            yield (inputs, outputs)
            line_pointer = line_pointer + self.batch_size 


def prediction_layer(x):
    # x.shape = (?,6040,5)
    x_cumsum = K.cumsum(x, axis=2)
    # x_cumsum.shape = (?,6040,5)

    output = K.softmax(x_cumsum)
    # output = (?,6040,5)
    return output


def prediction_output_shape(input_shape):

    return input_shape


def d_layer(x):

    return K.sum(x, axis=1)


def d_output_shape(input_shape):

    return (input_shape[0], )


def D_layer(x):

    return K.sum(x, axis=1)


def D_output_shape(input_shape):

    return (input_shape[0],)


def rating_cost_lambda_func(args):
    alpha=0.01 #in the paper they reported alpha = 0.01 and std = 1.0. THis is what was used in the repo.
    std=1.0
    pred_score, true_ratings, input_masks, output_masks, D, d = args
    pred_score_cum = K.cumsum(pred_score, axis=2)
    prob_item_ratings = K.softmax(pred_score_cum)
    accu_prob_1N = K.cumsum(prob_item_ratings, axis=2)
    accu_prob_N1 = K.cumsum(prob_item_ratings[:, :, ::-1], axis=2)[:, :, ::-1]
    mask1N = K.cumsum(true_ratings[:, :, ::-1], axis=2)[:, :, ::-1]
    maskN1 = K.cumsum(true_ratings, axis=2)
    cost_ordinal_1N = -K.sum((K.log(prob_item_ratings) - K.log(accu_prob_1N)) * mask1N, axis=2)
    cost_ordinal_N1 = -K.sum((K.log(prob_item_ratings) - K.log(accu_prob_N1)) * maskN1, axis=2)
    cost_ordinal = cost_ordinal_1N + cost_ordinal_N1
    nll_item_ratings = K.sum(-(true_ratings * K.log(prob_item_ratings)), axis=2)
    nll = std * K.sum(nll_item_ratings, axis=1) * 1.0 * D / (D - d + 1e-6) \
        + alpha * K.sum(cost_ordinal, axis=1) * 1.0 * D / (D - d + 1e-6)
    cost = K.mean(nll)
    cost = K.expand_dims(cost, 0)

    return cost

Functions

def D_layer(x)

Expand source code

def D_layer(x):

    return K.sum(x, axis=1)

def D_output_shape(input_shape)

Expand source code

def D_output_shape(input_shape):

    return (input_shape[0],)

def d_layer(x)

Expand source code

def d_layer(x):

    return K.sum(x, axis=1)

def d_output_shape(input_shape)

Expand source code

def d_output_shape(input_shape):

    return (input_shape[0], )

def prediction_layer(x)

Expand source code

def prediction_layer(x):
    # x.shape = (?,6040,5)
    x_cumsum = K.cumsum(x, axis=2)
    # x_cumsum.shape = (?,6040,5)

    output = K.softmax(x_cumsum)
    # output = (?,6040,5)
    return output

def prediction_output_shape(input_shape)

Expand source code

def prediction_output_shape(input_shape):

    return input_shape

def rating_cost_lambda_func(args)

Expand source code

def rating_cost_lambda_func(args):
    alpha=0.01 #in the paper they reported alpha = 0.01 and std = 1.0. THis is what was used in the repo.
    std=1.0
    pred_score, true_ratings, input_masks, output_masks, D, d = args
    pred_score_cum = K.cumsum(pred_score, axis=2)
    prob_item_ratings = K.softmax(pred_score_cum)
    accu_prob_1N = K.cumsum(prob_item_ratings, axis=2)
    accu_prob_N1 = K.cumsum(prob_item_ratings[:, :, ::-1], axis=2)[:, :, ::-1]
    mask1N = K.cumsum(true_ratings[:, :, ::-1], axis=2)[:, :, ::-1]
    maskN1 = K.cumsum(true_ratings, axis=2)
    cost_ordinal_1N = -K.sum((K.log(prob_item_ratings) - K.log(accu_prob_1N)) * mask1N, axis=2)
    cost_ordinal_N1 = -K.sum((K.log(prob_item_ratings) - K.log(accu_prob_N1)) * maskN1, axis=2)
    cost_ordinal = cost_ordinal_1N + cost_ordinal_N1
    nll_item_ratings = K.sum(-(true_ratings * K.log(prob_item_ratings)), axis=2)
    nll = std * K.sum(nll_item_ratings, axis=1) * 1.0 * D / (D - d + 1e-6) \
        + alpha * K.sum(cost_ordinal, axis=1) * 1.0 * D / (D - d + 1e-6)
    cost = K.mean(nll)
    cost = K.expand_dims(cost, 0)

    return cost

Classes

class DataSet (ratings_df, num_users, num_items, batch_size, rating_bucket, mode)

A datagenerator the feeds data in batches.

ratings_df: rating matrix, num_iters * num_users, entry is input rating rounded to integer batch_size: int, batch size, default is 64 num_users: int, number of users num_items: int, number of items mode: int, 0 for train, 1 for eval, 2 for test

Expand source code

class DataSet(Callback):
    """
    A datagenerator the feeds data in batches.

    ratings_df: rating matrix, num_iters * num_users, entry is input rating rounded to integer
    batch_size: int, batch size, default is 64
    num_users: int, number of users
    num_items: int, number of items
    mode: int, 0 for train, 1 for eval, 2 for test
    """
    def __init__(self,ratings_df,
        num_users,
        num_items,
        batch_size,
        rating_bucket,
        mode):

        self.num_users = num_users
        self.num_items = num_items
        self.batch_size = batch_size
        self.ratings_df = ratings_df
        self.rating_bucket = rating_bucket
        self.mode = mode

    def generate(self, eval=False):
        """
        a generator function yields ratings_df for each batch

        """
        line_pointer = 0
        while True:
            next_n_data_lines = list(islice(self.ratings_df, line_pointer, line_pointer+self.batch_size))
            if not next_n_data_lines:
                if self.mode == 0 and eval==False:
                    line_pointer = 0
                    next_n_data_lines = list(islice(self.ratings_df, line_pointer, line_pointer+self.batch_size))
                else:
                    break
            input_ranking_vectors = np.zeros((self.batch_size, self.num_users, self.rating_bucket), dtype='int8')
            output_ranking_vectors = np.zeros((self.batch_size, self.num_users, self.rating_bucket), dtype='int8')
            input_mask_vectors = np.zeros((self.batch_size, self.num_users), dtype='int8')
            output_mask_vectors = np.zeros((self.batch_size, self.num_users), dtype='int8')
            for i, line in enumerate(next_n_data_lines):
                user_ids = np.nonzero(line)[0]
                ratings_line = line[line != 0]

                if self.mode == 0 and len(user_ids) != 0:
                    # a random ordered list 0 to len(user_ids)-1
    
                    ordering = np.random.permutation(np.arange(len(user_ids)))
                    random_num = np.random.randint(0, len(ordering))
                    flag_in = (ordering < random_num)
                    flag_out = (ordering >= random_num)
                  
                    input_mask_vectors[i][user_ids] = flag_in
                    output_mask_vectors[i][user_ids] = flag_out

                    for j, (user_id, value) in enumerate(zip(user_ids, ratings_line)):
                        if flag_in[j]:
                            input_ranking_vectors[i, user_id, (value-1)] = 1
                        else:
                            output_ranking_vectors[i, user_id, (value-1)] = 1
                if self.mode == 2:
                    for j, (user_id, value) in enumerate(zip(user_ids, ratings_line)):
                        input_ranking_vectors[i, user_id, (value-1)] = 1                  

            inputs = {
                'input_ratings': input_ranking_vectors,
                'output_ratings': output_ranking_vectors,
                'input_masks': input_mask_vectors,
                'output_masks': output_mask_vectors}

            outputs = {'nade_loss': np.zeros([self.batch_size])}
            yield (inputs, outputs)
            line_pointer = line_pointer + self.batch_size 

Ancestors

• keras.callbacks.callbacks.Callback

Methods

def generate(self, eval=False)

a generator function yields ratings_df for each batch

Expand source code

def generate(self, eval=False):
    """
    a generator function yields ratings_df for each batch

    """
    line_pointer = 0
    while True:
        next_n_data_lines = list(islice(self.ratings_df, line_pointer, line_pointer+self.batch_size))
        if not next_n_data_lines:
            if self.mode == 0 and eval==False:
                line_pointer = 0
                next_n_data_lines = list(islice(self.ratings_df, line_pointer, line_pointer+self.batch_size))
            else:
                break
        input_ranking_vectors = np.zeros((self.batch_size, self.num_users, self.rating_bucket), dtype='int8')
        output_ranking_vectors = np.zeros((self.batch_size, self.num_users, self.rating_bucket), dtype='int8')
        input_mask_vectors = np.zeros((self.batch_size, self.num_users), dtype='int8')
        output_mask_vectors = np.zeros((self.batch_size, self.num_users), dtype='int8')
        for i, line in enumerate(next_n_data_lines):
            user_ids = np.nonzero(line)[0]
            ratings_line = line[line != 0]

            if self.mode == 0 and len(user_ids) != 0:
                # a random ordered list 0 to len(user_ids)-1

                ordering = np.random.permutation(np.arange(len(user_ids)))
                random_num = np.random.randint(0, len(ordering))
                flag_in = (ordering < random_num)
                flag_out = (ordering >= random_num)
              
                input_mask_vectors[i][user_ids] = flag_in
                output_mask_vectors[i][user_ids] = flag_out

                for j, (user_id, value) in enumerate(zip(user_ids, ratings_line)):
                    if flag_in[j]:
                        input_ranking_vectors[i, user_id, (value-1)] = 1
                    else:
                        output_ranking_vectors[i, user_id, (value-1)] = 1
            if self.mode == 2:
                for j, (user_id, value) in enumerate(zip(user_ids, ratings_line)):
                    input_ranking_vectors[i, user_id, (value-1)] = 1                  

        inputs = {
            'input_ratings': input_ranking_vectors,
            'output_ratings': output_ranking_vectors,
            'input_masks': input_mask_vectors,
            'output_masks': output_mask_vectors}

        outputs = {'nade_loss': np.zeros([self.batch_size])}
        yield (inputs, outputs)
        line_pointer = line_pointer + self.batch_size