import tensorflow as tf
import numpy as np
from command_line_args import arg_parser
from data import data_from_args
from model_helpers import build_model, init_weights, set_seeds
args = arg_parser.parse_args()
data = data_from_args(args)
NUM_HIDDEN = args.num_hidden or 100
set_seeds(123)
# Initialize the weights.
w_h = init_weights([data.num_inputs, NUM_HIDDEN])
w_o = init_weights([NUM_HIDDEN, 4])
b_h = init_weights([NUM_HIDDEN])
b_o = init_weights([4])
def model(X, w_h, w_o, b_h, b_o):
h = tf.nn.relu(tf.matmul(X, w_h) + b_h)
return tf.matmul(h, w_o) + b_o
py_x = model(data.X, w_h, w_o, b_h, b_o)
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(py_x, data.Y))
train_op = tf.train.RMSPropOptimizer(learning_rate=0.0008, decay=0.8, momentum=0.4).minimize(cost)
predict_op = tf.argmax(py_x, 1)
import numpy as np
from command_line_args import arg_parser
from data import data_from_args
from model_helpers import build_model, init_weights, set_seeds
args = arg_parser.parse_args()
data = data_from_args(args)
NUM_HIDDEN1 = args.num_hidden or 100
NUM_HIDDEN2 = args.num_hidden2 or 100
set_seeds(123)
# Initialize the weights.
w_h1 = init_weights([data.num_inputs, NUM_HIDDEN1])
w_h2 = init_weights([NUM_HIDDEN1, NUM_HIDDEN2])
w_o = init_weights([NUM_HIDDEN2, 4])
b_h1 = init_weights([NUM_HIDDEN1])
b_h2 = init_weights([NUM_HIDDEN2])
b_o = init_weights([4])
def model(X, w_h1, w_h2, w_o, b_h1, b_h2, b_o, keep_prob):
h1 = tf.nn.dropout(tf.nn.relu(tf.matmul(X, w_h1) + b_h1), keep_prob)
h2 = tf.nn.relu(tf.matmul(h1, w_h2) + b_h2)
return tf.matmul(h2, w_o) + b_o
def py_x(keep_prob):
import numpy as np
from command_line_args import arg_parser
from data import data_from_args
from model_helpers import build_model, init_weights, set_seeds
args = arg_parser.parse_args()
data = data_from_args(args)
NUM_HIDDEN1 = args.num_hidden or 100
NUM_HIDDEN2 = args.num_hidden2 or 100
set_seeds(123)
# Initialize the weights.
w_h1 = init_weights([data.num_inputs, NUM_HIDDEN1])
w_h2 = init_weights([NUM_HIDDEN1, NUM_HIDDEN2])
w_o = init_weights([NUM_HIDDEN2, 4])
b_h1 = init_weights([NUM_HIDDEN1])
b_h2 = init_weights([NUM_HIDDEN2])
b_o = init_weights([4])
def model(X, w_h1, w_h2, w_o, b_h1, b_h2, b_o, keep_prob):
if keep_prob < 1.0:
h1 = tf.nn.dropout(tf.nn.relu(tf.matmul(X, w_h1) + b_h1), keep_prob)
else:
h1 = tf.nn.relu(tf.matmul(X, w_h1) + b_h1)
h2 = tf.nn.relu(tf.matmul(h1, w_h2) + b_h2)
return tf.matmul(h2, w_o) + b_o
# logistic regression model
# https://en.wikipedia.org/wiki/Logistic_regression
import tensorflow as tf
import numpy as np
from command_line_args import arg_parser
from data import data_from_args
from model_helpers import build_model, init_weights, set_seeds
set_seeds(123)
args = arg_parser.parse_args()
data = data_from_args(args)
w = init_weights([data.num_inputs, 4])
b = init_weights([4])
def model(X, w, b):
return tf.matmul(X, w) + b
# Predict y given x using the model.
py_x = model(data.X, w, b)
# We'll train our model by minimizing a cost function.
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(py_x, data.Y))
train_op = tf.train.GradientDescentOptimizer(0.05).minimize(cost)
# And we'll make predictions by choosing the largest output.
# Linear Regression model
# https://en.wikipedia.org/wiki/Linear_regression
import tensorflow as tf
import numpy as np
from command_line_args import arg_parser
from data import data_from_args
from model_helpers import build_model, init_weights, set_seeds, make_output
set_seeds(123)
args = arg_parser.parse_args()
data = data_from_args(args)
w = init_weights([data.num_inputs, 4])
b = init_weights([4])
def model(X, w, b):
return tf.matmul(X, w) + b
# Predict y given x using the model.
py_x = model(data.X, w, b)
# We'll train our model by minimizing a cost function.
cost = tf.reduce_mean(tf.pow(py_x - data.Y, 2))
train_op = tf.train.GradientDescentOptimizer(0.05).minimize(cost)
# And we'll make predictions by choosing the largest output.
predict_op = tf.argmax(py_x, 1)