220 likes | 420 Views
Case-based Reasoning. Introduction. What is CBR? Case-based reasoning is a problem solving paradigm. CBR also is an approach to incremental, sustained learning. When a problem is successfully solved When an attempt to solve a problem fails as well.
E N D
Introduction What is CBR? • Case-based reasoning is a problem solving paradigm. • CBR also is an approach to incremental, sustained learning. • When a problem is successfullysolved • When an attempt to solve a problem failsas well.
Central tasks of all case-based reasoningmodels • Identify the current problem situation. • Find a past case similar to the new one. • Use that case to suggest a solution to the current problem. • Evaluate the proposed solution, and update the system by learning from this experience.
1. Representation of Cases • Deciding what to store in a case. • Finding an appropriate structure. • Deciding how the case memory should be organized . • How to integrate the case memory structure into a model of general domain knowledge.
2. Case Retrieval • Identify Feature • Initially Match • Select
3. Case Reuse • Copy • Adapt • Transformational reuse • Derivational reuse
4. Case Revision • Evaluate solution • Repair fault
5.Case Retainment Learning • Extract • Index • Integrate
Case-based planning • Case-based planning is the reuse of past successful plans in order to solve new planning problems. • Extracting behavioral knowledge from an expert in form of individual cases. This knowledge can be reused to achieve the specific open goals in the current plan.
On-line Case-based Plan Adaptation • Allows the AI engine to create and adapt plans in an efficient and effective manner while performing the task.
Case-Based Planning in WARGUSDarmok system • Darmok was designed to play the game of WARGUS and apply the plan adaptation techniques. • Darmok learns behaviors from expert demonstrations, and then uses case-based planning to play the game reusing the learnt behaviors.
Plan Representation Language • A behavior has two main parts: • The declarative part • The procedural part • The state of a behavior • Goal
On-Line Case-Based Plan Adaptation • Plan Dependency Graph Generation • Precondition-success condition matcher (ps-matcher). • Removal of unnecessary actions • Adaptation for unsatisfied preconditions
Plan Dependency Graph • Taking the success conditions of the root of the plan. • Finding out which actions in the plan contribute to the achievement of these success conditions. • These actions are called direct sub-plans for the sub-goal. • 4- All direct sub-plans are added to the plan dependency graph. Then the plan dependency graph generator analyzes the preconditions of each of these direct sub-plans.