1 / 9

Lecture 28 Modeling (3): Expert System and Reinforcement Learning

Lecture 28 Modeling (3): Expert System and Reinforcement Learning. Outline. ANN and Knowledge Based Expert System (KBES) A Brief Introduction to rule-based KBES Integrating KBES and ANN. Knowledge Based Expert System.

lowell
Download Presentation

Lecture 28 Modeling (3): Expert System and Reinforcement Learning

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Lecture 28Modeling (3):Expert System and Reinforcement Learning

  2. Outline • ANN and Knowledge Based Expert System (KBES) • A Brief Introduction to rule-based KBES • Integrating KBES and ANN (C) 2001-2003 by Yu Hen Hu

  3. Knowledge Based Expert System • A Knowlege Based Expert System (KBES) is a Computer Program which solves a specific type of problems by codifying human experts knowledge in a knowledge base, and by mimicking the human problem solving process. (C) 2001-2003 by Yu Hen Hu

  4. KBES (Cont'd) • An KBES consists of a database which stores problem dependent facts, a human interface to communicate with user, and two special modules: Knowledge Base – a special data base where human experts' knowledge is codified. Inference Engine – a mechnism which enables the utilization of knowledge and facts to solve a given problem. (C) 2001-2003 by Yu Hen Hu

  5. KBES Structure • The Knowledge base consists of "RULES" : Rule 0021: IF (1) QRS  .11sec.on any two limb leads AND (2) Sd.  .04sec. on lead I or aVL AND (3) terminal R present lead Vl THEN (a) QRS .11 seconds; AND (b) terminal QRS rightward and anterior; AND (c) incomplete right bundle branch block • Rules are elicited from human experts' knowledge. • Inference engine provides search and binding mechanism to perform "logic reasoning" – A process in which a chain of rules are invoked (fired) based on given facts. (C) 2001-2003 by Yu Hen Hu

  6. ANN and KBES Any rule based expert system can be realized with a MLP network with threshold activation function to implement AND, OR, and NOT logic operations: (C) 2001-2003 by Yu Hen Hu

  7. ANN and Rules MLP implementation of Rule 0021. The hidden layer performs AND operation, the output layer performs OR operation. (C) 2001-2003 by Yu Hen Hu

  8. EXtract KBEs from An ANN • KBNN (Knowledge Based Neural Network) – A rule based KBES can be used to initialize a MLP. Then its performance can be fine-tuned using learning. For example, if a trained MLP has the following form: What will be the corresponding rules? (C) 2001-2003 by Yu Hen Hu

  9. Extracting Rules • One possible rule derived directly from the hyperplane: If 0.8•a + 0.3•b – 0.9•c + 1 > 0, then z is TRUE Not quite a comprehensible rule! • Rule Extraction Approaches: Eliminating weights with small magnitude (pruning) Training network with quantized weights (–1, 0, +1) N_of_M rule extraction – Determine whether an OR gate or an AND gate should be used. (C) 2001-2003 by Yu Hen Hu

More Related