def test_pairwise_transform(self):
X = np.array([[0.2, 0.3, 0.4], [0.1, 0.7, 0.4], [0.3, 0.4, 0.1],
[0.8, 0.4, 0.3], [0.9, 0.35, 0.25]])
y = np.array([0, 1, 0, 0, 2])
qid = np.array([1, 1, 1, 2, 2])
ranker = LambdaRankNN(input_size=X.shape[1],
hidden_layer_sizes=(
16,
8,
),
activation=(
'relu',
'relu',
),
solver='adam')
X1, X2, Y, weight = ranker._transform_pairwise(X, y, qid)
self.assertTrue((X1 == np.array([[0.1, 0.7, 0.4], [0.1, 0.7, 0.4],
[0.8, 0.4, 0.3]])).all())
self.assertTrue((X2 == np.array([[0.2, 0.3, 0.4], [0.3, 0.4, 0.1],
[0.9, 0.35, 0.25]])).all())
self.assertTrue((Y == np.array([1, 1, 0])).all())
self.assertTrue((weight == np.array(
[0.36907024642854247, 0.13092975357145753,
0.36907024642854247])).all())
def lambdarank(entity_info,flag):
if entity_info:
if flag == 0:
X = []
print(entity_info)
for i in entity_info:
X.append(i[2:])
X = np.array(X)
print(X)
ranker = LambdaRankNN(input_size=X.shape[1], hidden_layer_sizes=(16,8,), activation=('relu', 'relu',), solver='adam')
ranker.model.load_weights('LambdaRankNNmaster/Examples/ccks20_entity.h5')
y_pred = ranker.predict(X)
K.clear_session()
return y_pred
else:
X = []
print(entity_info)
for i in entity_info:
X.append(i[5:])
X = np.array(X)
print(X)
ranker = LambdaRankNN(input_size=X.shape[1], hidden_layer_sizes=(16,8,), activation=('relu', 'relu',), solver='adam')
ranker.model.load_weights('LambdaRankNNmaster/Examples/ccks20_relation.h5')
y_pred = ranker.predict(X)
K.clear_session()
return y_pred
else:
return entity_info
def lambdarank(entity_info,flag):
if flag == 0:
X = []
print(entity_info)
for i in entity_info:
X.append(i[2:])
X = np.array(X)
print(X)
ranker = LambdaRankNN(input_size=X.shape[1], hidden_layer_sizes=(16,8,), activation=('relu', 'relu',), solver='adam')
ranker.model.load_weights('LambdaRankNNmaster\Examples\entity_select225.h5')
y_pred = ranker.predict(X)
return y_pred
else:
X = []
print(entity_info)
for i in entity_info:
X.append(i[4:])
X = np.array(X)
print(X)
ranker = LambdaRankNN(input_size=X.shape[1], hidden_layer_sizes=(16,8,), activation=('relu', 'relu',), solver='adam')
ranker.model.load_weights('LambdaRankNNmaster\Examples\drelation_select22515.h5')
y_pred = ranker.predict(X)
return y_pred
import numpy as np
from LambdaRankNN import LambdaRankNN, RankNetNN
import h5py
import keras
# generate query data
X = np.array([[0.2, 0.3, 0.4], [0.1, 0.7, 0.4], [0.3, 0.4, 0.1],
[0.8, 0.4, 0.3], [0.9, 0.35, 0.25]])
y = np.array([3, 1, 2, 1, 2])
qid = np.array([1, 1, 1, 2, 2])
ranker = LambdaRankNN(input_size=X.shape[1],
hidden_layer_sizes=(
16,
8,
),
activation=(
'relu',
'relu',
),
solver='adam')
ranker.fit(X, y, qid, epochs=5)
#ranker.model.save_weights('ranker_weights.h5')
y_pred = ranker.predict(X)
print(y_pred)
ranker.evaluate(X, y, qid, eval_at=2)
30000]
test_X, test_y, test_queries = test_X[:
30000, :], test_y[:
30000], test_queries[:
30000]
print("Train Shape", train_X.shape, train_y.shape)
print("Valid Shape", vali_X.shape, vali_y.shape)
print("Test Shape", test_X.shape, test_y.shape)
# Train LambdaRankNN model
ranker = LambdaRankNN(input_size=train_X.shape[1],
hidden_layer_sizes=(
8,
4,
),
activation=(
'relu',
'relu',
),
solver='adam')
ranker.fit(train_X, train_y, train_queries, epochs=10)
train_y_pred = ranker.predict(train_X)
vali_y_pred = ranker.predict(vali_X)
test_y_pred = ranker.predict(test_X)
ranker.evaluate(test_X, test_y, test_queries, eval_at=2)
df1 = pd.DataFrame(train_y_pred)
df2 = pd.DataFrame(vali_y_pred)
df3 = pd.DataFrame(test_y_pred)
Created on Sun Apr 19 00:04:15 2020
@author: Rishab
"""
from LambdaRankNN import LambdaRankNN
from DataLoader import read_dataset
X, y, queries = read_dataset('./data_output/train.tsv')
VX, Vy, Vqueries = read_dataset('./data_output/test.tsv')
y = y.astype(int)
qids = queries.astype(int)
ranker = LambdaRankNN(input_size=X.shape[1],
hidden_layer_sizes=(
16,
8,
),
activation=(
'relu',
'relu',
),
solver='adam')
ranker.fit(X, y, qids, epochs=20)
# y_pred = ranker.predict(TX)
Vy = Vy.astype(int)
Vqids = Vqueries.astype(int)
ranker.evaluate(VX, Vy, Vqids, eval_at=2)