from src.main.model.regression.linearRegression import LinearRegression
from src.main.model.model import Config, Parameters
import tensorflow as tf
from src.main.dataset.inputdata import regression_data
from src.main.model.functions.regularization import Ridge
# model configurations
config = Config(feature_num=2, batch_size=500, learning_rate=0.01, epoche=500)
# dataset
dataset = regression_data(TRUE_W=[[17.0], [4]],
TRUE_b=4,
NUM_EXAMPLES=10000,
batch_size=1000)
# implemented model
sess = tf.Session()
lr = LinearRegression(dataset=dataset, config=config, parameters=Parameters())
sess.run(tf.global_variables_initializer())
lr.training(session=sess)
print(sess.run(lr.weights))
model.final_state,
model.optimization], feed_dict={feature: X, target: Y, init_state: training_state})
training_loss += training_loss_
if step % 100 == 0 and step > 0:
if verbose:
print("Average loss at step", step, "for last 250 steps:",
training_loss / 100)
training_losses.append(training_loss / 100)
training_loss = 0
return training_losses
if __name__ == '__main__':
configuration = Config(feature_num=1,
batch_size=batch_size,
epoche=20,
learning_rate=learning_rate,
min_learning_rate=learning_rate,
num_layers=1,
num_unrollings=1,
num_classes=num_classes,
num_nodes=1,
dropout=0.1,
state_size=state_size,
num_steps=num_steps)
training_losses = main(config=configuration,
num_epochs=1,
num_steps=num_steps)
plt.plot(training_losses)
plt.show()
from src.main.model.classification.logisticRegression import LogisticRegression from src.main.model.model import Config, Parameters import tensorflow as tf from src.main.dataset.inputdata import classification_data # model configurations config = Config(feature_num=10, batch_size=50, learning_rate=0.01) # dataset NUM_EXAMPLES=10000 dataset = classification_data(batch_size=1000, n_samples=NUM_EXAMPLES, n_features=10) # model sess = tf.Session() model = LogisticRegression(dataset=dataset, config=config, parameters=Parameters()) sess.run(tf.global_variables_initializer()) model.training(session=sess) print(sess.run([model.weights, model.bias])) print(sess.run(model.accuracy))