eustin

NUM_CONTROL_USERS = 6 # total number of users in control group

control_users = np.zeros(NUM_CONTROL_USERS)

#> array([0., 0., 0., 0., 0., 0.])

eustin

rng = np.random.default_rng(1337) # random number generator using the one true seed
sns.set_style('white')

eustin

import random
import numpy as np
from statsmodels.stats.proportion import score_test_proportions_2indep
from scipy import stats
from numba import njit, prange
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
import seaborn as sns

eustin

https://upload.wikimedia.org/wikipedia/commons/c/c5/%21%21%21_Mdina_buildings_05.jpg
https://upload.wikimedia.org/wikipedia/commons/d/d5/%21-2011-debowa-leka-palac-abri.jpg
https://upload.wikimedia.org/wikipedia/commons/c/c2/%21-20100523-szlichtyngowa-wiatrak-abri.jpg
https://upload.wikimedia.org/wikipedia/commons/c/ca/%21Illegal_street_ads_on_a_bicycle.jpg
https://upload.wikimedia.org/wikipedia/commons/f/fc/%21-2010-hetmanice-kosciol-abri.jpg
https://upload.wikimedia.org/wikipedia/commons/2/27/%21-2011-debowa-leka-kosciol-abri.jpg
https://upload.wikimedia.org/wikipedia/commons/0/0b/%22%2Barya%2B%22_penjual_jamu_tradisional_-_%EA%A6%A2%EA%A6%BA%EA%A6%B4%EA%A6%AD%EA%A7%80%EA%A6%A2%EA%A6%BA%EA%A6%B4%EA%A6%AD%EA%A6%A4%EA%A7%80_%EA%A6%97%EA%A6%A9%EA%A6%B8_%EA%A6%A0%EA%A6%BF%EA%A6%A2%EA%A6%B6%EA%A6%B1%EA%A6%B6%EA%A6%AA%EA%A6%BA%EA%A6%B4%EA%A6%A4%EA%A6%AD%EA%A7%80_Pilangsari_2019.jpg
https://upload.wikimedia.org/wikipedia/commons/c/cd/%22..._Although_a_dress_uniform_is_not_a_part_of_the_regular_equipment%2C_most_of

eustin

hist = model.fit(
    [query_embeddings, docs_averaged_embeddings], 
    relevance_grades_prob_dist, 
    epochs=50, 
    verbose=False
)

eustin

query_input = tf.keras.layers.Input(shape=(1, EMBEDDING_DIMS, ), dtype=tf.float32, name='query')
docs_input = tf.keras.layers.Input(shape=(NUM_DOCS_PER_QUERY, EMBEDDING_DIMS, ), dtype=tf.float32, 
                name='docs')

expand_batch = ExpandBatchLayer(name='expand_batch')
dense_1 = tf.keras.layers.Dense(units=3, activation='linear', name='dense_1')
dense_out = tf.keras.layers.Dense(units=1, activation='linear', name='scores')
scores_prob_dist = tf.keras.layers.Dense(units=NUM_DOCS_PER_QUERY, activation='softmax', 
                      name='scores_prob_dist')

eustin

class ExpandBatchLayer(tf.keras.layers.Layer):
    def __init__(self, **kwargs):
        super(ExpandBatchLayer, self).__init__(**kwargs)
        
    def call(self, input):
        queries, docs = input
        batch, num_docs, embedding_dims = tf.unstack(tf.shape(docs))
        expanded_queries = tf.gather(queries, tf.zeros([num_docs], tf.int32), axis=1)
        return tf.concat([expanded_queries, docs], axis=-1)

eustin

NUM_DOCS_PER_QUERY = 5
EMBEDDING_DIMS = 2

eustin

batch_loss = tf.reduce_mean(per_example_loss)

print(batch_loss)

eustin

per_example_loss = tf.reduce_sum(
    relevance_grades_prob_dist * tf.math.log(relevance_grades_prob_dist / scores_prob_dist),
    axis=-1
)

print(per_example_loss)