[PYTHON] Japanese translation of public teaching materials for Deep learning nanodegree

** Updated from time to time **: Links will be added from the translated ones.

1.First of all

I have completed the Deep learning nanodegree [^ name] that I have been taking since March 2017. This course is charged, but high quality Jupyter Notebook is open to the public for free, so I decided to translate it into Japanese after reviewing [^ Target]. The GitHub repository is here. Udacity has permission to publish.

[^ Name]: I don't care, but sometimes I officially call it Deep learning ** foundation ** nanodegree, and sometimes I call it Deep learning nanodegree ** foundation **, so I didn't know which was correct. For the time being, this article follows Introduction to Udacity Deep Learning Nanodegree. If you know the official name, I would appreciate it if you could tell me.

[^ Target]: The latest version as of July 13, 2017 is the target of translation.

It's a super-free translation, but I hope it will be helpful for those who are wondering about taking the course. Also, I would be grateful if you could point out any mistranslations.

2. Deep learning nanodegree

This is a paid course for Udacity. It covers all the basic topics of deep learning, from implementing Neural networks with NumPy to Convolutional neural networks, Recurrent neural networks, and Generative adversarial networks. Please refer to some wonderful articles about this course.

• Introduction to Udacity Deep Learning Nanodegree
• Impressions of completing the Udacity Deep Learning Foundation

Some of the teaching materials for this course are available on youtube and GitHub. In this project, we will translate the notebook published on GitHub into Japanese [^ manpower].

[^ Manual]: Ideally, the Deep learning algorithm should be implemented and machine translated, but this time it is not powerful and manual. We are looking for people who can cooperate!

3. Table of contents

Notebooks are roughly divided into tutorials and projects. Answers are provided for the tutorial exercises, but no answers are provided for the project exercises. This is because the latter is an assignment for Udacity students. Therefore, in this project, the tutorial, which is completed as one notebook, is the target of translation.

We classify them into 5 groups at our discretion as follows. () Is the original title.

3.1 Introduction

• Introduction to TensorFlow (Intro to TensorFlow)
• Reputation analysis with TFLearn ([Sentiment analysis with TFLearn](https:: //github.com/haltaro/deep-learning-in-japanese/blob/master/intro-to-tflearn/TFLearn_Sentiment_Analysis_Solution.ipynb)): TensorFlow wrapper TFLearn, movie I will challenge the reputation analysis of the review.
• Handwritten number recognition with TFLearn and MNIST
• Weight initialization to improve performance

3.2 Convolutional neural networks

• Transfer learning with ConvNet

3.3 Recurrent neural networks

• Introduction to RNN (Character-wise RNN)
• Word2Vec（Embeddings）
• Reputation analysis with RNN ([Sentiment analysis with RNN](https: //) github.com/haltaro/deep-learning-in-japanese/blob/master/sentiment-rnn/Sentiment_RNN_Solution.ipynb))): Implementing a long short term memory-based recurrent neural network in TensorFlow for reputation analysis of movie reviews I will challenge.
• Seq2Seq (Sequence to sequence deep-learning-in-japanese / blob / master / seq2seq / sequence_to_sequence_implementation.ipynb))): Implement Seq2Seq in TensorFlow and implement a model to sort input character strings.

3.4 Generative adversarial networks

• Autoencoder
• GAN (Generative Adversatial Network on MNIST) at MNIST
• Deep Convolutional GAN（Deep Convolutional GAN）

3.5 Other

• Introduction to Tensorboard (Tensorboard)
• Reinforcement learning with DQN ([Deep Q-learning](https:: //github.com/haltaro/deep-learning-in-japanese/blob/master/reinforcement/Q-learning-cart.ipynb))): Implement Deep Q-learning network in TensorFlow and Cat-Pole of OpenAI Gym I will challenge the game.
• Batch normalization（Batch normalization）

4. Environment construction

The requirements.txt in each directory shows the minimum environment required to run the notebook.

pip

You can build the environment with pip3 install -r requirements.

Conda

There is an environment file for Conda in the ʻenvironments` directory. Please use the file suitable for your platform.

License MIT License Copyright (c) 2017 Udacity, Inc.