LoginSignup
6
6

More than 5 years have passed since last update.

2017新人教育スケジュール(機械学習関連)

Last updated at Posted at 2017-07-12

1週目

6/12 (配属初日) :v:

業務紹介

githubの登録

機械学習

  • AIとMLとDLの関係性
  • 頻度統計 v.s. ベイズ統計
    • 最尤推定 v.s. ベイズ推定
  • 過学習(overfitting)
    • 対策: 正則化、dropout
  • パラメータチューニング
    • train-validation-test-split
  • 教師有モデル
    • 分類モデル
      • NN (Neural Network)
        • gradient descent
        • back propagation
      • SVM (Support Vector Machine)
        • まだ知らなくて良い:kernel trick
      • DT (Decision Tree)
        • まだ知らなくて良い:informational gain
      • RF (Random Forest)
        • 参考:bootstrap resampling

Python

  • jupyter notebookの紹介
test.ipynb
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np

x = np.arange(0,10,0.1)
y = np.sin(x)
plt.plot(x,y)

6/13 :v:

scikit-learn

  • sklearnのワークフロー
    1. model = Model()
    2. model.fit()
    3. model.predict()

6/14 :v:

tensorflow

  • tensorflowのワークフロー
  • インストール w/ anaconda @ windows 10

Deep Neural Network

  • DNNの基本的な話
  • activations
    • tanh
    • relu
    • lrelu
    • prelu
    • elu
  • optimizer
    • sgd
    • momentum
    • adagrad/ adadelta/ rmsprop
    • adam
  • initializer
    • xavier init.
    • he init.

6/15 :v:

keras

  • kerasのワークフロー

Autoencoder

  • denoising AE
  • sparse AE
  • variational AE (またあとで説明

Anomaly Detection

6/16 :v:

Convolutional Neural Network

  • convolution
  • pooling

Transfer Learning

Normalization

  • batch normalization
  • weight normalization
  • layer normalizaiton

Image Recognition

  • LeNet
  • AlexNet
  • ZFNet
  • VGGNet
  • NIN
  • GoogLeNet
  • ResNet

2週目

6/19 :v:

Keras

  • Practice: CNN

6/20 :v:

Keras

  • Practice: Transfer Learning

6/21 :v:

Keras

  • Practice: Transfer Learning

Model Compression

  • Deep Compression
  • Depse-wise Conv.
    • Xception
    • Mobilenet

6/22 :v:

Keras

  • Practice: Transfer Learning

Similarity Learning

  • Siamese Network

6/23 :v:

Object Detection

  • R-CNN
  • SPPNet
  • Fast R-CNN
  • Faster R-CNN
  • Overfeat
  • AttentionNet
  • SSD
  • YOLO

3週目

6/26 :v:

Semantic Segmentation

  • Deconvolution
  • FCN
  • Mask R-CNN

Recurrent Neural Networks

  • LSTM
  • GRU
  • Bi-RNN

Neural Machine Translation

  • Seq2Seq
  • Attention
  • Slicenet

Image Caption Generation

Keras

  • Practice: RNN-LSTM

6/27 :v:

Natural Language Processing

  • word2vec
    • Skipgram
    • CBoW(Continuous Bag of Words)
    • Glove
  • sentence2vec
    • BoW
    • Recursive Neural Network
    • Glove

6/28 :v:

Generative Models

  • GAN: Generative Adversarial Networks
    • DCGAN
  • Conditional GAN
    • Pix2Pix

Adversarial Examples

6/29 :v:

Generative Models

  • VAE: Variational Autoencoder
    • DRAW
  • Pixel RNN/ Pixel CNN
    • Wavenet
    • Bytenet

6/30 :v:

Understanding CNN

  • Maximally Activating Patches
  • Occlusion Maps
  • Saliency Maps
  • Features Inversion via backprop

Neural Style Transfer

  • Deep Dream
  • Neural Art

7/3 :v:

Reinforcement Learning

  • MDP
  • Policy Gradient
    • AlphaGo

7/12 以降

Reinforcement Learning

  • Q-Learning
    • DQN
  • Actor-Critic
    • A3C

NN w/ External Memory

  • Neural Turing Machine
  • Differential Neural Computer

Meta Learning

  • Evolutional Strategy

One-Shot Learning

  • Matching Networks

Semi-Supervised Learning

  • GAN / VAE / Virtual Adversarial Training

Self-Driving Cars

  • NVIDIA Model
  • Localization
  • Sensor Fusion
  • Kalman filter
  • PID Control
  • SLAM
  • Path Planning

Robotics

  • Perception / Decision / Action
6
6
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
6
6