securetorobert

# Defaults to a tf.float32 scalar.
numeric_feature_column = tf.feature_column.numeric_column(key="SepalLength")

# Represent a tf.float64 scalar.
numeric_feature_column = tf.feature_column.numeric_column(key="SepalLength",
                                                          dtype=tf.float64)

# Represent a 10-element vector in which each cell contains a tf.float32.
vector_feature_column = tf.feature_column.numeric_column(key="Bowling",
                                                         shape=10)

securetorobert

import pandas as pd
import numpy as np
import tensorflow as tf

from sklearn.model_selection import train_test_split

#load training data
train_df = pd.read_csv('train.csv', index_col='ID')

#split into features and target

securetorobert

def input_fn(X, y, batch_size=16, epochs=1, shuffle=False):
  #create a dictionary of features and their values
  features = {key:np.array(value) for key,value in dict(X).items()} 
  #create a tf.data compliant dataset
  d = tf.data.Dataset.from_tensor_slices((features, y))
  
  #we need our data in batches
  d = d.batch(batch_size).repeat(epochs)
  
  #optionally shuffle our data

securetorobert

X.info()

feature_columns = [
    tf.feature_column.numeric_column('crim', dtype=tf.float64),
    tf.feature_column.numeric_column('zn', dtype=tf.float64),
    tf.feature_column.numeric_column('indus', dtype=tf.float64),
    tf.feature_column.numeric_column('chas', dtype=tf.int64),
    tf.feature_column.numeric_column('nox', dtype=tf.float64),
    tf.feature_column.numeric_column('rm', dtype=tf.float64),
    tf.feature_column.numeric_column('age', dtype=tf.float64),

securetorobert

optimizer = tf.train.FtrlOptimizer(learning_rate=0.01, l1_regularization_strength=0.1)

estimator = tf.estimator.LinearRegressor(feature_columns=feature_columns, optimizer=optimizer)

estimator.train(input_fn= lambda: input_fn(X_train, y_train), steps=20)

securetorobert

evaluate_result = estimator.evaluate(input_fn=lambda: input_fn(X_train, y_train))
print("Training results")
for key in evaluate_result:
   print("   {}, was: {}".format(key, evaluate_result[key]))
 
evaluate_result = estimator.evaluate(input_fn=lambda: input_fn(X_test, y_test))
print("Training results")
for key in evaluate_result:
   print("   {}, was: {}".format(key, evaluate_result[key]))

securetorobert

preds = estimator.predict(input_fn=lambda: input_fn(X_test, y_test))

predictions = np.array([item['predictions'][0] for item in preds])

securetorobert

import tensorflow as tf
import tensorflow_hub as hub
import pandas as pd
import numpy as np
import shutil

import seaborn as sns
import matplotlib.pyplot as plt

MODULE_SPEC_50 = 'https://tfhub.dev/google/nnlm-en-dim50/1'

securetorobert

def make_input_fn(df, num_epochs):
  
  return tf.estimator.inputs.pandas_input_fn(
    x = df,
    y = df['sentiment'],
    batch_size = 128,
    num_epochs = num_epochs,
    shuffle = True,
    queue_capacity = 1000,
    num_threads = 1

securetorobert

CSV_COLUMNS = ['medv', 'crim', 'zn', 'indus', 'chas', 'nox', 'rm', 'age', 'dis', 'rad', 'tax', 'ptratio', 'black', 'lstat']
LABEL_COLUMN = 'medv'
DEFAULTS = [[0.0], [0.0], [0.0], [0.0], [0], [0.0], [0.0], [0.0], [0.0], [0], [0], [0.0], [0.0], [0.0]]

def read_dataset(filename, mode, batch_size = 16):
  def _input_fn():
    def decode_csv(value_column):
      columns = tf.decode_csv(value_column, record_defaults = DEFAULTS)
      features = dict(zip(CSV_COLUMNS, columns))
      label = features.pop(LABEL_COLUMN)