Lesson two: Getting Started with deep learning
Lecturer: David (data analysis engineer)
This course mainly introduces the basic principles of many neural networks, very very basic understanding.
0, Mind map preview:
One, deep neural network
1. How neurons work
--This is a biological neuron, which is then abstracted from it and made into a m-p neuron pattern.
2. Introduction to Neural networks
--1943 M-p Neuron model
--1956 Sensing Machine
--1986 Distributed Representation
by Hinton (the father of neural networks?) ) Proposed
-- 9 combinations of 6 neurons .
Red Dog
Neuron color Object |
Red |
Yellow |
Blue |
Car |
Red car |
Yellow car |
Blue car |
Horse |
Red horse |
Huang Ma |
Horse |
Dog | TD valign= "Top" width= "142" >
Yellow Dog |
Blue Dog |
In the beginning, 9 neurons were needed to represent these combinations, and then after the distributed representation, 6 neurons could be used , and by their 22 combinations, 9 combinations were realized. This method.
--1986 Inverse propagation algorithm
--1994 long and short memory network
--2006 Deep Neural Network
--2007 convolutional Neural network
3. Why do you learn so much in depth now?
--"Big" data
At present, the technology development is better, the network has rich data.
Deep learning: It takes a lot of data to train his abilities.
--"Deep" model
The computing power of the current computer is strong.
4. Neural network classification
--Feedforward Neural network
--Deep neural network (full-link)
--Optimizing deep neural networks
TensorFlow (more popular),torch ,Theano ,Caffe ,mxnet , pytorch
-- test:http://playground.tensorflow.org A tool test for optimizing deep neural networks
--convolutional neural networks
--dealing with the problem of image recognition
--Cyclic neural network (RNN)
--long and short Memory network (LSTM)
--Gate Loop Network (GRU)
--Production discriminant network
Second, the application of deep learning
1. Image recognition
2. Language recognition
3. Machine translation
4. Image generation
Third, how to learn deep learning
1. Mathematics
--Linear algebra
--Calculus
--Convex Optimization calculation method
--Probability theory, etc.
2. Machine learning
3. Programming
--Algorithms and data structures
--python
4. Deep learning
--Related Materials recommended:
Deep learning,Lan Goodfellow, Yoshua ,Bengio, etc.
--Paper website recommendation: Domain Name:arXiv
--Recommendation of relevant public courses
CS231N & Hinton
Python Learning (ii)--Introduction to deep learning