A new way to amplify the case-based reasoning (CBR) knowledge-base using randomization. This method allows knowledge amplification without deteriorating the CBR's resolution time and it was applied for the route planning domain.
1. Knowledge-Based Randomization for
Amplification
Authors:
• Prof. Thouraya Bouabana-Tebibel (LCSI, ESI)
• Prof. Stuart H. Rubin (SPAWAR)
• Dr. Lydia Bouzar-Benlabiod (LCSI, ESI)
• Miled Basma BENTAIBA-LAGRID (LCSI, ESI)
• Maria Roumeissa Hanini (LCSI, ESI)
2. 2
Introduction
The concept of
problem solving for
humans
The case-based
reasoning imitates the
human thinking
An evolutionary case-
base versus a static
case-base
4. 4
Definitions
Case / Case-Base
• Problem solutionCase
• Descriptors componentsForm of the case
•Example 𝐶1: ,ܣ ,ܤ ܥ → ܺ, 𝑌
5. How to increase
accuracy and
efficiency of CBR’s
problem resolution?
Current Solutions
• Feed the case-base
using inference
methods
Problem
• A massive case-base
may deteriorate the
CBR’s rapidity of search
Proposed Solution
• Amplify the knowledge
using randomization
Problem
• The generated cases
may not be valid
Solution
Validate the generated
cases before their use
5
Motivation
6. • A generic method was proposed (S. H. Rubin)
6
Traveling
Robots
Scheduling
systems
Refrigerator
design
Cyber context
Military
strategies
Related Work
Randomization
8. Proposed Approach
Problem Substitution Rule 1
• Let C1: prb1 sol1 (primary case)
• And C2: prb2 sol2 (pair substitute case)
• prb2 may substitute prb1 if it is resolved by sol1 sol2 = sol1
• The substitution of prb2 by prb1 must be performed in the same
context substitution case includes prb1 and sol1
• Example:
• Case1: A, B, C X,Y (primary case)
• Case2: D, E X,Y (pair substitute case)
• Case4: A, B, C, F, G X,Y,V, Z (substitution case)
• Result: Case4’: D, E, F, G X,Y,V, Z
9. Proposed Approach
Solution Substitution Rule 2
• Let C1: prb1 sol1 (primary case)
• And C2: prb2 sol2 (pair substitute case)
• sol2 may substitute sol1 if it is also a solution for prb1 prb2 =prb1
• The substitution of sol2 by prb1 must be performed in the same
context substitution case includes prb1 and sol1
• Example:
• Case5: D, E, F, G U (primary case)
• Case4’: D, E, F, G X,Y,V, Z (pair substitute case)
• Case7: D, E, F, G, H, K U,T (substitution case)
• Result: Case7’: D, E, F, G, H, K X,Y,V, Z,T
10. Application Domain
Formalization
• Problem solutionCase
• SP, {waypoints (Pi)}, EP routes: SPpj,…, pkEPForm of the case
• C1: p1, p2, p12 p1p2, p2p9, p9p10, p10p19, p19p11, p11p12,
p12p13, p13p13
Example
19. Conclusion
Main contributions are:
(1) Randomization technique for amplification,
(2) Segmentation of the case-base,
(3) New method for route planning based on the CBR
(4) Domain specific validation
21. Knowledge-Based Randomization for
Amplification
Authors:
• Prof. Thouraya Bouabana-Tebibel t_tebibel@esi.dz
• Prof. Stuart H. Rubin stuart.rubin@navy.mil
• Dr. Lydia Bouzar-Benlabiod l_bouzar@esi.dz
• Miled Basma BENTAIBA-LAGRID bm_bentaiba@esi.dz
• Maria Roumeissa Hanini m_hanini@esi.dz
Editor's Notes
Humans can solve problems that they confront on daily basis. One of the methods of doing that is using previous experiences and adapt their solutions to the newly encountered problems.
CBR imitates the human thinking. It has a case-base where a case is referred to an experience captured from the real word.
It is clear that when we have a static and non-evolving case-base the problem resolution process won’t be efficient. On the other hand, a massive case-base may slow down the resolution process. This is the aim of using randomization process to amplify the case-base without affecting the resolution time. Don’t worry, I will explain what the randomization is in next slides.
First defined by (G. Chaitin 1975) it means that information or knowledge can be effectively compressed until that representation of the compressed information is random; or in other words, pattern-less.
Randomization was first used for knowledge amplification purposes by professor Tebibel. The randomization allows us to amplify the knowledge, and also keeping the case-base optimized.
Randomization is domain specific
A generic method was proposed (S. H. Rubin)
The formalization of the problem may vary from a domain to another
Randomization has been used for different purposes, not only for knowledge amplification. In our work, we are focusing on knowledge amplification using randomization
An approach for knowledge induction based on randomization is presented in [7]. The approach consists in deriving knowledge by exploiting analogies among the base cases. We propose, in this paper, novel types of analogies. The whole of analogies behind the carried-out substitutions are given in the two following subsections, which respectively deal with the problem substitution and solution substitution. Both types of substitution require three specific cases, called trio. The trio is composed of a primary case paired with a substitute case and associated with a substitution case.
Also, for short, we write in what follows a case as:
Casei: prbi→ soli
In problem substitution, we define case pairing as follows. Paired cases are a pair of primary and substitute cases which have different problem parts and solutions which are either the same or whose one belongs to the other.
In solution substitution, we define case pairing as follows: Paired cases are a pair of primary and substitute cases, which have different solution parts and problems which are either the same or whose one belongs to the other.
We must ensure that any point in the problem part is included in a route in the solution part
We must replace the solution part with another one that starts and ends with the same points to ensure the continuity
Two types of segments: paired cases of problem substitution PS and paired cases for solution substitution are saved in the SS segments.
Each segment is represented by a delegate which is a generic description of the problem parts if we are talking about PS or a generic description for the solution part if we are talking about SS
The delegate is composed of every descriptor (component) and the number of its occurrence. If we project it to the route planning problem, the delegate is for each point in the problem part we store the number of its occurrence in the PS segment, and for each route we store the number of its occurrence in the SS segment.
Each new generated case is compared to the delegates, and if the similarity is higher that the threshold we store it in the segment and pair it with every possible case in the segment.
The cases are not physically inserted in the segments, but their reference is.
Percentage of number of generated cases per number of initial cases, augmented by 140% and 100%
100 random problems, and test how much from 100 the system can resolve. It’s about 44%