본문 바로가기

Enginius/Robotics

Machine Learning - Stanford Lecture


Lecture 1 Machine Learning (Stanford)

 - http://www.youtube.com/watch?v=UzxYlbK2c7E

 Machine Learning approached from diverse fields. 

 Inter-disciplinary topic

 having large impact // lots of implication // to science and industry

 
 Early work in AI - viewed as a new capability for computers. 
 - reading hand-written letters, translating - extract characters
 - fly a helicopter 
 : using learning algorithm. pretty much the only approach. 

 - database mining. 
 - checks -> processed by learning algorithm / ?? 
 - credit card // fraud, stolen check
 - guess what clients wants 

 state of art machine learning algorithm 

 Using Matlab, Octave. 

Machine Learning definition 
 Arthur Samuel: Field of study that gives computers the ability to learn w/o being explicitly programmed. 
  - The computer learns how to play Checker game
 
 Tom Mitchell: Well-posed Learning Problem: A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improved with experience E. 

1. Supervised Learning
 Cartesian coordinate - predicting based on events accumulated 
 - Regression. (A: continuous)
 what about discrete data set? 
 - Classification (discrete values: 0, 1)
 what about diverse input data? what if its infinite? 

2. Learning Theory
 reading zip code?
 will it work? how many training data will be needed? 

3. Unsupervised Learning
 unlike supervised learning that data gives right answers, 
 just a data set // having no right answers..? 'classification'? grouping 
 image processing, -> vision, image processing (clustering)
 

















 

'Enginius > Robotics' 카테고리의 다른 글

노인과 로봇  (0) 2011.10.15
Rat Brain Robot  (0) 2011.10.06
e-puck H/W  (0) 2011.08.09
Robot Simulator - BSA (Backtracking Spiral Algorithm)  (0) 2011.02.13
Robot Simulator - Non grid, Divie and Conquer, Spiral  (0) 2011.02.10