[PYTHON] Matters examined at the study session in February

I did machine learning (Scikit-learn, TensorFlow) at Mooc (Udacity, Coursera).

Mooc

• [Neural Networks for Machine Learning | Coursera](https://www.coursera.org/learn/ neural-networks/home/welcome)

• Deep Learning - Udacity

Andrew Ng is famous, but Google's Deep Learning --Udacity is nice to handle with Scikit-learn or TensorFlow.

Neural Networks for Machine Learning

• Mr. Hinton
• The execution environment was Octave
• I gave up and installed it, but this time it's over

Deep Learning - Udacity

• It's great that assignments are given in Jupyter Notebook (it works!)
• The theoretical aspect may be unsatisfactory
• I recommend it because there are some scenes that say "Oh, I'm having a hard time"

Deep Learning - Udacity

• Do tensorflow / tensorflow / examples / udacity at master · tensorflow / tensorflow

Setup

• Since it was troublesome to prepare the environment, I used Docker

• TensorFlow v1.0 was released, so I updated it

• I want to update the Docker image to the latest in a batch --Qiita

Scikit-learn

• The first task was "Let's go back to Logistic using Scikit-learn"

• tensorflow/1_notmnist.ipynb at master · tensorflow/tensorflow

• Machine Learning, etc: notMNIST dataset

Scikit-learn

• scikit-learn/scikit-learn: scikit-learn: machine learning in Python

• sklearn.linear_model.LogisticRegression — scikit-learn 0.18.1 documentation

• L1 Penalty and Sparsity in Logistic Regression — scikit-learn 0.18.1 documentation

Python: Extracting common elements

• How to compare list and list and retrieve common elements in list --Qiita

v = [1,2,3]
w = [2,3,4]
[x for x in v if x in w] # [2,3]

• Detailed python comprehension --Qiita

TensorFlow

It was about performing the same task in Tensorflow.

• tensorflow/2_fullyconnected.ipynb at master · tensorflow/tensorflow

• Assignment 2 : building a 3-hidden layer network? - Courses / Deep Learning - Udacity Discussion Forum

• Assignment 2: 2 hidden layers and NaN loss - Courses / Deep Learning - Udacity Discussion Forum

• Deep MNIST for Experts  |  TensorFlow

I was satisfied because I made a 3-layer NN like this and achieved accuracy 94.2%. When I extract a part of the code, it looks like this.

def weight_variable(shape):
  initial = tf.truncated_normal(shape, stddev=1.732/sum(shape))
  return tf.Variable(initial)

def bias_variable(shape):
  initial = tf.constant(0.1, shape=shape)
  return tf.Variable(initial)
  
batch_size = 128
hidden_layer_size = 1024
input_layer_size = 28*28
output_layer_size = 10

graph = tf.Graph()
with graph.as_default():
    # Input data. For the training data, we use a placeholder that will be fed
    # at run time with a training minibatch.
    tf_train_dataset = tf.placeholder(tf.float32, shape=(batch_size, image_size * image_size))
    tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))
    tf_valid_dataset = tf.constant(valid_dataset)
    tf_test_dataset = tf.constant(test_dataset)
    
    # Variables
    weight1 = weight_variable( (input_layer_size, hidden_layer_size) )
    bias1 = bias_variable( [hidden_layer_size] )
    
    # Hidden Layer
    hidden_layer = tf.nn.relu(tf.matmul(tf_train_dataset, weight1) + bias1)
    
    # Variables
    weight2 = weight_variable( (hidden_layer_size, output_layer_size) )
    bias2 = bias_variable( [output_layer_size] )
    
    logits = tf.matmul(hidden_layer, weight2) + bias2
    loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=tf_train_labels))
    
    # optimizer
    optimizer = tf.train.AdamOptimizer(0.001).minimize(loss)
    # optimizer = tf.train.GradientDescentOptimizer(0.5).minimize(loss)
    
    # prediction
    train_prediction = tf.nn.softmax(logits)
    valid_prediction = tf.nn.softmax(tf.matmul(tf.nn.relu(tf.matmul(tf_valid_dataset, weight1) + bias1), weight2) + bias2)
    test_prediction = tf.nn.softmax(tf.matmul(tf.nn.relu(tf.matmul(tf_test_dataset, weight1) + bias1), weight2) + bias2)