Scribd
Upload
106 views31 pages

TensorFlow Overview for Deep Learning

This document provides an overview of TensorFlow and deep learning. It begins by explaining machine learning and how deep learning is a method for implementing artificial intelligence by allowing machines to learn features and make predictions from data without being explicitly programmed. It then discusses what TensorFlow is, describing it as a library released by Google for quickly building neural networks and noting that Python is required. It also mentions using a GPU can aid in deep learning tasks. Finally, it outlines next steps as training and testing neural networks on data while addressing issues like overfitting through hyperparameters, loss functions, and regularization techniques.

Uploaded by

Syed Adeel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
106 views31 pages

TensorFlow Overview for Deep Learning

This document provides an overview of TensorFlow and deep learning. It begins by explaining machine learning and how deep learning is a method for implementing artificial intelligence by allowing machines to learn features and make predictions from data without being explicitly programmed. It then discusses what TensorFlow is, describing it as a library released by Google for quickly building neural networks and noting that Python is required. It also mentions using a GPU can aid in deep learning tasks. Finally, it outlines next steps as training and testing neural networks on data while addressing issues like overfitting through hyperparameters, loss functions, and regularization techniques.

Uploaded by

Syed Adeel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 31

An Overview of

Tensorflow + Deep learning

沒一村
About me
• 邱弘毅（沒一村）
• 交通大學電資學士班 3 年級
• 沒一村前端筆記 & 沒一村生活點滴
• JavaScript lover, R language, python
• Full stack developer, React Native
• 目前專注於 Deep learning 相關應用
• https://www.facebook.com/noootown
Before we start
• 程式碼：
• https://github.com/noootown/Forex-DQN

• 受用的話，請給我一顆 star :P
Outline
• What is Machine Learning?
• What is Deep learning?
• What is Tensorflow?
• Do it!
• Summary
What is Machine learning ？
What is Machine learning ？

Field of study that gives computers the ability to learn

without being explicitly programmed

Arthur Lee Samuel, 1959


Machine Learning
實踐 AI 的一種手段

透過各種 Machine learning 演算法，從少量或大量的資料中學


出一個 rule 。藉這個 rule 對未來做預測。

問題分類： cluster, regression, classification AI

常見演算法 : KNN, SVM…… ML


pic from: http://www.csie.ntnu.edu.tw/~u91029/Classification.html
問題 評估 修正
Outline
• What is Machine Learning?
• What is Deep learning?
• What is Tensorflow?
• Do it!
• Summary
What is Deep learning ？
Deep Learning
也是實踐 AI 的一種手段

讓機器自己提取特徵，自己試誤，自己學習
AI
各種不同的 Model
ML
Ex: CNN, RNN, LSTM
DL
Deep Learning vs 類神經網路
模擬人類神經傳導機制
How Neuron works?
Control weight and bias
Fully connected
They need activation!!!
Ex: relu, sigmoid……
Target: Minimize loss

1 2
網路
3 4

1 3 Loss function
f( ) = Ex: Cross entropy,
3 5 mean square
How ? Gradient Decent
Optimize function
Learning rate
0<lr<1
影響收斂的速度
影響結果的好壞
Review
Gradient 受下列因素影響

Loss function
Learning rate
Opitimize function
Gradient descent 問題
Speed up?

Mini-batch

Stuck in Local minimum?

Momentum
Network structure
到底要幾層？每層要開多大？

依照 input 的大小和特性決定
一般來說第一層大一點，越來越小
Outline
• What is Machine Learning?
• What is Deep learning?
• What is Tensorflow?
• Do it!
• Summary
Tensorflow
Google 釋出內部使用的 library

快速建造需要的神經網路

You need Python!

Maybe you demand a nice GPU like 1080Ti


Tensorboard
Scalars

Graphs

Distribution

Histogram
Outline
• What is Machine Learning?
• What is Deep learning?
• What is Tensorflow?
• Do it!
• Summary
Outline
• What is Machine Learning?
• What is Deep learning?
• What is Tensorflow?
• Do it!
• Summary
Summary & Way to go
• Mathemetics
Training Testing
• Hyperparameter 的過程 data
data
• 克服 overfitting 的方法
• Vanish Gradient
• ……
Loss Function Early Stopping

Learning Rate Regularization

Activation Function Dropout

Optimize Function

You might also like