PR for review on slim version.

Brain_Integ_Y.csv

@@ -0,0 +1 @@
+Censored,Survival0,1440,3931,4700,2110,6910,10240,3940,4240,3180,4690,4050,3840,7171,13221,14051,2180,12290,6320,8140,4140,4660,330,3800,1500,3821,9321,1810,5060,1331,9580,3570,1080,2540,1381,2681,2731,2600,830,1141,1871,1390,1591,2370,1640,450,36670,2241,2531,3911,1451,471,1451,1510,5480,1110,4630,3330,10620,4420,4540,620,410,880,8800,5430,1190,7721,2940,1480,1240,2130,2020,5230,2440,5750,1440,3680,540,4280,5110,4551,9531,2861,4521,1671,2411,710,1540,5050,3000,7370,12330,9140,7530,4800,771,370,5440,7130,4680,1420,3860,5150,1130,3830,5851,2720,2900,300,1350,3600,60,4600,3790,1051,6361,1321,4361,2280,1380,940,1110,2790,1000,820,5190,1650,2540,1470,1530,1140,7270,10480,5671,11010,491,35740,13350,11061,18281,12221,81,16310,1940,5380,8141,15681,14281,15191,14941,14261,37331,24931,22891,14581,22180,15471,27721,13541,11121,23811,14121,5731,9191,14211,12101,11391,10211,22191,9621,7360,7751,9161,1821,9551,8351,8461,8890,4381,5631,3681,3421,5441,9351,4920,6071,9681,19431,14011,7481,12791,8461,5231,5231,7061,5481,4431,5310,2051,4071,2571,2071,55461,5761,5321,4611,64230,15850,6820,14910,22860,14010,20000,370,26600,2140,3540,40680,6050,5120,28750,34701,2421,730,3490,19151,47520,46950,4560,32000,1990,2690,35711,27610,9331,8000,3150,2421,28690,5760,18860,7881,18340,7091,12771,840,1550,7220,4441,6291,6511,5821,5161,4711,4311,10121,9081,9641,7431,7961,7600,10331,7771,6561,6780,3470,8141,5220,6480,5471,25651,15670,2280,13510,11831,170,15250,20520,12510,24330,17620,40840,23790,29070,39780,28350,1110,5920,4920,6480,7270,15780,4660,9870,11200,44450,3980,22350,9841,12011,71,61,14530,7751,15880,11201,5711,4541,12941,12571,4671,5441,6861,6271,6111,8621,4161,4331,4871,6111,4551,6851,9921,11641,3260,5591,5691,7211,4491,6551,4571,6771,8681,3281,11151,6221,5671,1741,31,2030,3510,12201,41,11,5031,1141,5301,1991,7381,1941,12271,22871,29181,32531,52551,1221,1531,9081,7051,37251,1391,5260,961,6711,13991,17061,17520,71,4341,51,4941,231,13591,1841,13001,31,9641,4551,31,31,901,5331,2101,4421,12011,5851,11891,71,15401,151,7181,11301,101,41,1121,1621,10790,1131,3431,9561,501,4941,2860,19331,30,11521,10041,3721,1901,31,141,1691,4191,71,9001,10401,4781,7921,41,6151,21,31,821,5151,551,4911,5121,6511,5331,741,3201,3361,31,17211,13820,12621,13871,20781,21071,10780,12091,12171,12941,7580,9620,3881,6300,5781,13011,4381,01,01,71,2041,2431,721,701,2871,1791,2421,2311,391,71,3171,4031,3541,3331,2741,4141,2921,2301,3721,5841,1051,721,1341,771,3841,4821,4271,5881,4511,4421,4140,3721,4081,2791,4421,4971,5091,581,5021,5081,3371,3951,27021,28601,9931,9391,1661,3011,631,5960,18911,26501,5711,12451,10760,7421,7851,9921,10691,29181,22891,41131,26021,11731,8051,5671,4281,10990,2411,7151,6081,6091,2491,621,5641,8260,51661,9140,9611,15001,5661,4541,6501,5991,4861,13411,10261,7951,7871,-11,18680,8210,9541,13971,13141,4111,30,2401,11911,7141,11271,4871,7061,15531,13931,14261,1147

data_providers.py

@@ -0,0 +1,52 @@
+# -*- coding: utf-8 -*-
+
+import numpy as np
+import pandas as pd
+
+
+DATA_FILE = ['./Brain_Integ_X.csv', './Brain_Integ_Y.csv']
+
+def data_providers(data_file=['./Brain_Integ_X.csv', './Brain_Integ_Y.csv']):
+    """The function for reading datasets.
+
+    This function reads the features, lable datasets separately and filters out
+    only the dead patients' respective observations for further processings.
+
+  Args:
+    data_file: list of strings representing the paths for input files.
+    Here the input features and input ground truth values are provided separately.
+
+  Returns:
+    `Numpy array`, extracted feature matrix and label column.
+
+  Example:
+    >>> read_dataset()
+    ( [[2.3, 2.4, 6.5],[2.3, 5.4,3.3]], [12, 82] )
+
+    """
+
+    data_feed = pd.read_csv(data_file[0], skiprows=[0], header=None)
+    labels_feed = pd.read_csv(data_file[1], skiprows=[1], header=0)
+    survival = labels_feed['Survival']
+    censored = labels_feed['Censored']
+
+    survival = survival.values
+    censored = censored.values
+    data = data_feed.values
+    data = np.float32(data)
+
+    #Filtering only the dead patients for survival analysis
+    censored_survival = survival[censored == 1]
+    censored_data = data[censored == 1]
+
+    y = np.asarray(censored_survival)
+    x = np.asarray(censored_data)
+
+    print('Shape of X : ', x.shape)
+    print('Shape of Y : ', y.shape)
+    return (x, y)
+
+
+if __name__ == '__main__':
+
+    data_x, data_y = data_providers(DATA_FILE)

dnn_model.py

@@ -0,0 +1,42 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import absolute_import, division, print_function
+import numpy as np
+import tensorflow as tf
+
+
+slim = tf.contrib.slim
+
+
+def multilayer_nn_model(inputs, hidden_layers, n_classes, beta,
+                        scope="deep_regression_model"):
+    """Creates a deep regression model.
+
+    This function takes input as the required parameters to build a deep
+    neural network and builds the layer-wise network. Once its instance is
+    called by parsing the input feedings, this function will perform the
+    feed-forward and returns the output layer responses.
+
+    Args:
+        inputs: A node that yields a `Tensor` of size [total_observations,
+        input_features].
+
+    Returns:
+        predictions: `Tensor` of shape (1) (scalar) of response.
+        end_points: A dict of end points representing the hidden layers.
+    """
+
+    with tf.variable_scope(scope, 'deep_regression', [inputs]):
+        end_points = {}
+        with slim.arg_scope([slim.fully_connected],
+                            activation_fn=tf.nn.relu,
+                            weights_regularizer=slim.l2_regularizer(beta)):
+            net = slim.stack(inputs,
+                             slim.fully_connected,
+                             hidden_layers,
+                             scope='fc')
+            end_points['fc'] = net
+            predictions = slim.fully_connected(net, n_classes, activation_fn=None,
+                                               scope='prediction')
+            end_points['out'] = predictions
+            return predictions, end_points

read_record.py

@@ -0,0 +1,57 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import absolute_import, division, print_function
+import tensorflow as tf
+import numpy as np
+
+
+TFRECORD_FILE = "data.tfrecords"
+
+
+def read_tfrecord(tfrecords_filename="data.tfrecords"):
+    """The function for reading the tf.record file.
+
+    This function takes input as the protobuff file name. Then, given the
+    header strings along with their data-types, this function will
+    retrieve those respective values from the byte file through a loop.
+    Finaaly, it will convert those strings back to the desired data-type
+    (here its np.asarray[dtype=np.float32]).
+
+    Args:
+        tfrecords_filename: This carries the tfrecord file-name with the file-path.
+
+    Returns:
+        predictors: A numpy array(M*N) contains 'np.float32' elements.
+        This variable returns the feature values matrix.
+        gnd_truths: A numpy array(M*1) contains 'np.float32' elements.
+        This variable returns the observed survival values vector.
+
+    """
+
+    predictors = []
+    gnd_truths = []
+    record_iterator = tf.python_io.tf_record_iterator(path=TFRECORD_FILE)
+
+    for element in record_iterator:
+        example = tf.train.Example()
+        example.ParseFromString(element)
+
+        predictor_string = (example.features.feature['predictor_string']
+                            .bytes_list
+                            .value[0])
+        gnd_truth_string = (example.features.feature['gnd_truth_string']
+                            .bytes_list
+                            .value[0])
+
+        predictor = np.fromstring(predictor_string, dtype=np.float32)
+        gnd_truth = np.fromstring(gnd_truth_string, dtype=np.float32)
+
+        predictors.append((predictor))
+        gnd_truths.append((gnd_truth))
+
+    return predictors, gnd_truths
+
+
+if __name__ == '__main__':
+
+    predictors, gnd_truths = read_tfrecord(TFRECORD_FILE)

read_record_demo.py

@@ -0,0 +1,63 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import absolute_import, division, print_function
+import tensorflow as tf
+import numpy as np
+
+
+TFRECORD_FILE = "data.tfrecords"
+
+
+def read_tfrecord(tfrecords_filename="data.tfrecords"):
+    """The function for reading the tf.record file.
+
+    This function takes input as the protobuff file name. Then, given the
+    header strings along with their data-types, this function will
+    retrieve those respective values from the byte file through a loop.
+    Finaaly, it will convert those strings back to the desired data-type
+    (here its np.asarray[dtype=np.float32]).
+
+    Args:
+        tfrecords_filename: This carries the tfrecord file-name with the file-path.
+
+    Returns:
+        predictors: A numpy array(M*N) contains 'np.float32' elements.
+        This variable returns the feature values matrix.
+        gnd_truths: A numpy array(M*1) contains 'np.float32' elements.
+        This variable returns the observed survival values vector.
+
+    """
+
+    predictors = []
+    gnd_truths = []
+    record_iterator = tf.python_io.tf_record_iterator(path=TFRECORD_FILE)
+
+    for element in record_iterator:
+        example = tf.train.Example()
+        example.ParseFromString(element)
+
+        predictor_string = (example.features.feature['predictor_string']
+                            .bytes_list
+                            .value[0])
+        gnd_truth_string = (example.features.feature['gnd_truth_string']
+                            .bytes_list
+                            .value[0])
+
+        predictor = np.fromstring(predictor_string, dtype=np.float32)
+        gnd_truth = np.fromstring(gnd_truth_string, dtype=np.float32)
+
+        print('predictor : ', predictor)
+        print('gnd_truth : ', gnd_truth)
+
+        predictors.append((predictor))
+        gnd_truths.append((gnd_truth))
+
+    return predictors, gnd_truths
+
+
+if __name__ == '__main__':
+
+    predictors, gnd_truths = read_tfrecord(TFRECORD_FILE)
+
+    # print('predictors np.array : ', predictors)
+    # print('gnd_truth np.array : ', gnd_truths)