Machine Learning.pptx
- 2. Introduction Computers learning from medical records to identify which treatments are most effective for new diseases. Houses learning from experience to optimize energy costs based on the particular usage patterns of their occupants. Personal software assistants learning the evolving interests of their users in order to highlight relevant stories from online morning newspaper. If we know how to make computer learn it will explore new uses of computers. Detailed understanding of information processing algorithms for machine learning might lead to better understanding of human learning abilities or disabilities.
- 3. Contd… Still there are complexity in identifying how to make computers to learn nearly equal to human. Algorithms have been invented for certain type of learning tasks. Speech Recognition Data Mining Loan applications, Financial Transactions, Medical Records Some of the applications (Where Machine Learning used) Predict recovery rates of pneumonia patients Detect fraudulent credit cards drive autonomous vehicles on public highways play games
- 4. Contd… Theoretical results have been developed that characterize the fundamental relationship among the number of training examples observed The number of hypothesis under consideration the expected error in learned hypothesis Several recent applications of Machine Learning Learning to recognize spoken words Learning to drive autonomous vehicle Learning to classify new astronomical structures Learning to play world-class backgammon
- 5. Well posed Learning Problems Definition: A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, improves with experience E. Example 1: A Checkers Learning Problem: Task T: Playing Checkers Performance Measure P: percent of games won against opponents Training Experience E: Playing practice games against itself
- 6. Contd… Example 2: A handwriting recognition learning problem: Task T: recognizing and classifying handwritten words within images Performance Measure P: percent of words correctly classified Training Experience E: a database of handwritten words with given classifications Example 3: A robot driving learning problem: Task T: driving on public four-lane highways using vision sensors Performance Measure P: average distance traveled before an error (as judged by human overseer) Training Experience E: a sequence of images and steering commands recorded while observing a human driver
- 7. Contd… To summarize, the goal is to define precisely a class of problems that encompasses interesting forms of learning, to explore algorithms that solve such problems and to understand the fundamental structure of learning problems and processes.
- 8. Designing a Learning System
- 9. Designing a Learning System To illustrate the basic design issues and approaches of Machine Learning, let us consider designing a program to learn to play checkers with the goal of entering it in the world of checkers tournament. Choosing the Training Experience The first design choice is to choose the type of training experience from which our system will learn. The type of training experience available will have impact on the success or failure of the learner. One key attribute is whether the training experience provides direct or indirect feedback regarding the choices made by the performance of the system.
- 10. Direct
- 11. Indirect
- 12. Contd… Direct training examples consist of individual checkers board states and the correct move for each. Indirect information consisting of the move sequences and final outcomes of various games played. In this case, correctness of the specific move inferred indirectly from the fact that the game won or lost. Additionally the learner faces an problem of credit assignment for each move in the sequence with respect to final outcome. It is particularly difficult problem because the game can be lost even when early moves are optimal later followed by poor moves.
- 13. Contd… The Second important attribute of training experience is the degree to which the learner control the sequence of training examples. The learner may rely the teacher to select informative board states and to provide the correct move. Alternatively, the learner may itself create new moves and clarify with the teacher when learner finds some confusion. One more choice is the learner may learn by playing against itself with no teacher present. This will provide a choice to experiment with novel board states which has not yet considered which may be a most promising moves while play.
- 14. Contd… The third important attribute of training experience is how well it represents the distribution of examples over the final system performance P. Learning is most reliable when the training examples follow a distribution similar to that of future test examples. In checkers, the Performance Metric is the percent of games the system wins in the world tournament. If the training experience E only contains training experience gained from playing with itself only means it is an obvious danger. That means, it might not contains full representation of distribution of situations.
- 15. Contd… For example, learner might never encounter certain crucial board states that are likely played in human checkers champion. It is often necessary to learn from a distribution of examples that is somewhat different from those on which final system will be evaluated. But finding such kind of distributions is difficult and most theory of Machine learning is based on the assumption that the distribution of training examples is identical to the distribution of test examples. It is important to keep in mind that this assumption must often violated in practice. Let us decide that our system will train by playing games against itself which has an advantage of no external trainer need to present.
- 16. Contd… A Checkers learning problem: Task T: Playing Checkers Performance Measure P: percent of games won in the world tournament Training Experience E: games played against itself In order to complete the design of the learning system, we must now choose 1. The exact type of knowledge to be learned 2. A representation for this target knowledge 3. A learning mechanism
- 17. Choosing the target function The next design choice is to determine exactly what type of knowledge will be learned and how this will be used by the performance program. Checker playing program can generate the legal moves from any board state. The program should need to choose the best move from that state. This learning task is a representative of large class of tasks for which the best search strategy is not known.
- 18. Contd… ChooseMove : B M Reduce the problem of improving performance P at Task T into problem of learning particular function such as Choose Move Indirect Learning Experience ChooseMove V: B R R - Real Number Target function V assign high score to better states. System successfully learn target function V Generate the successor board states for each legal move using V and therefore the best legal move.
- 19. Contd… Target value V(b) for an arbitrary board state b in B as follows: 1. If b is a final board state that is won, then V(b) = 100. 2. If b is a final board state that is lost, then V(b) = -100. 3. If b is a final board state that is drawn, then V(b) = 0. 4. If b is not the final state in the game, then V(b) = V(b’), where b’ is the best final board state that can be achieved starting from b and playing optimally until the end of the game.