01. Introduction to Deep learning

1. What is a neural network

example 1-Single neural network

Housing price prediction

• linear regression problem

• ReLU (Rectified Linear Unit 修正线性单元) function
  

• A “Neuron” is what connects the input(x=size) with the output(y=price)

• 

example 2-Multiple neural network

• other features besides size, should also be taken into account
• input layer -> hidden unit->y
  

2. Supervised learning with Neural Networks

supervised learning-> given correct output matching input data

I. regression & classification problems

• regression
  • predict results within a continuous output->map input variables to some continuous function
• classification
  • discrete output->discrete categories

II. Types of Neural Network

• Concolution Neural Network(CNN)
  • images
• Recurrent Neural Network(RNN)
  • one-dimentional sequence data
  • e.g. translation, speech recognition
• hybrid neural network architecture
  • e.g. autonomous driving

III. Structured vs. unstructured data

• Structured data:
  • have well-defined features such as price, age
• Unstructured data:
  • e.g. pixel, raw audio, text

3. Why is Deep learning taking off

Due to

• a large amount of data available through the digitization of the society
• faster computation
• innovation in the development of neural network algorithm

4. (unwatched) Geoffrey Hinton interview