Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"
-
Upload
ieeeciscyprus -
Category
Documents
-
view
159 -
download
0
Transcript of Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"
![Page 1: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/1.jpg)
Scalability Improvement of Genetics-Based Machine Learning
to Large Data Sets
Hisao IshibuchiOsaka Prefecture University, Japan
Workshop on Grand Challenges of Computational Intelligence (Cyprus, September 14, 2012)
![Page 2: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/2.jpg)
Contents of This Presentation
Introduction1. Basic Idea of Evolutionary Computation2. Genetics-Based Machine Learning3. Parallel Distributed Implementation4. Computation Experiments5. Conclusion
![Page 3: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/3.jpg)
Contents of This Presentation
Introduction1. Basic Idea of Evolutionary Computation2. Genetics-Based Machine Learning3. Parallel Distributed Implementation4. Computation Experiments5. Conclusion
![Page 4: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/4.jpg)
Basic Idea of Evolutionary Computation
Environment
Population
Individual
![Page 5: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/5.jpg)
Basic Idea of Evolutionary Computation
Environment
Population
Individual
(1)
(1) Natural selection in a tough environment.
Environment
Population
Good Individual
![Page 6: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/6.jpg)
Basic Idea of Evolutionary Computation
Environment
Population
Individual
(1) (2)
Environment
Population
Environment
Population
New Individual
(1) Natural selection in a tough environment.(2) Reproduction of new individuals by crossover and mutation.
Good Individual
![Page 7: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/7.jpg)
Basic Idea of Evolutionary Computation
Environment
Population
Environment
Population
Iteration of the generation update many times(1) Natural selection in a tough environment.(2) Reproduction of new individuals by crossover and mutation.
![Page 8: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/8.jpg)
Applications of Evolutionary ComputationDesign of High Speed TrainsEnvironment
Population
Individual = Design ( )
![Page 9: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/9.jpg)
Applications of Evolutionary ComputationDesign of Stock Trading AlgorithmsEnvironment
Population
Individual = Trading Algorithm ( )
![Page 10: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/10.jpg)
Contents of This Presentation
Introduction1. Basic Idea of Evolutionary Computation2. Genetics-Based Machine Learning3. Parallel Distributed Implementation4. Computation Experiments5. Conclusion
![Page 11: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/11.jpg)
Genetics-Based Machine LearningKnowledge Extraction from Numerical Data
If (condition) then (result)If (condition) then (result)If (condition) then (result)If (condition) then (result)If (condition) then (result)
Training dataEvolutionary Computation
Design of Rule-Based Systems
![Page 12: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/12.jpg)
Environment
Population
Design of Rule-Based Systems
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 13: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/13.jpg)
Environment
Population
Design of Decision Trees
Individual = Decision Tree ( )
![Page 14: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/14.jpg)
Environment
Population
Design of Neural Networks
Individual = Neural Network ( )
![Page 15: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/15.jpg)
Multi-Objective EvolutionMinimization of Errors and Complexity
Erro
rSm
all
Larg
e
ComplexitySimple Complicated
Initial Population
![Page 16: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/16.jpg)
Multi-Objective EvolutionMinimization of Errors and Complexity
Erro
rSm
all
Larg
e
ComplexitySimple Complicated
![Page 17: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/17.jpg)
Multi-Objective EvolutionA number of different neural networks
Erro
rSm
all
Larg
e
ComplexitySimple Complicated
![Page 18: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/18.jpg)
Multi-Objective EvolutionA number of different decision trees
Erro
rSm
all
Larg
e
ComplexitySimple Complicated
![Page 19: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/19.jpg)
Multi-Objective EvolutionA number of fuzzy rule-based systems
Erro
rSm
all
Larg
e
ComplexitySimple Complicated
![Page 20: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/20.jpg)
Contents of This Presentation
Introduction1. Basic Idea of Evolutionary Computation2. Genetics-Based Machine Learning3. Parallel Distributed Implementation4. Computation Experiments5. Conclusion
![Page 21: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/21.jpg)
Environment
Population
Difficulty in Applications to Large DataComputation Load for Fitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
Fitness Evaluation:Evaluation of each individual using the given training data
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 22: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/22.jpg)
Environment
Population
Difficulty in Applications to Large DataComputation Load for Fitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
Fitness Evaluation:Evaluation of each individual using the given training data
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 23: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/23.jpg)
EnvironmentPopulation
Difficulty in Applications to Large DataComputation Load for Fitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training dataFitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 24: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/24.jpg)
EnvironmentPopulation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training dataFitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
The Main Issue in This Presentation How to Decrease the Computation Load
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 25: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/25.jpg)
EnvironmentPopulation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training dataFitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Fitness CalculationStandard Non-Parallel Model
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 26: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/26.jpg)
EnvironmentPopulation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training dataFitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Fitness CalculationStandard Non-Parallel Model
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 27: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/27.jpg)
EnvironmentPopulation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
Fitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Fitness CalculationStandard Non-Parallel Model
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 28: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/28.jpg)
EnvironmentPopulation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
Fitness EvaluationIf ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Fitness CalculationStandard Non-Parallel Model
Individual = Rule-Based System ( )If ... Then ...If ... Then ...If ... Then ...If ... Then ...
![Page 29: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/29.jpg)
A Popular Approach for Speed-UpParallel Computation of Fitness Evaluation
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If we use n CPUs, the computation load for each CPU can be 1/n in comparison with the case of a single CPU (e.g., 25% by four CPUs)
CPU 1
CPU 2
CPU 3
CPU 4
Fitness Evaluation
![Page 30: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/30.jpg)
Another Approach for Speed-UpTraining Data Reduction
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If we use x% of the training data, the computation load can be reduced to x% in comparison with the use of all the training data.
Training data Subset
![Page 31: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/31.jpg)
Another Approach for Speed-UpTraining Data Reduction
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If we use x% of the training data, the computation load can be reduced to x% in comparison with the use of all the training data.
Training data Subset
![Page 32: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/32.jpg)
Another Approach for Speed-UpTraining Data Reduction
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If we use x% of the training data, the computation load can be reduced to x% in comparison with the use of all the training data.
Training data Subset
![Page 33: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/33.jpg)
Another Approach for Speed-UpTraining Data Reduction
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If we use x% of the training data, the computation load can be reduced to x% in comparison with the use of all the training data.
Training data Subset
![Page 34: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/34.jpg)
Another Approach for Speed-UpTraining Data Reduction
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Training data
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Difficulty: How to choose a training data subsetThe population will overfit to the selected training data subset.
Training data Subset
![Page 35: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/35.jpg)
Idea of Windowing in J. Bacardit et al.: Speeding-up Pittsburgh learning classifier systems: Modeling time and accuracy. PPSN 2004.
Training data
Training data are divided into data subsets.
![Page 36: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/36.jpg)
Idea of Windowing in J. Bacardit et al.: Speeding-up Pittsburgh learning classifier systems: Modeling time and accuracy. PPSN 2004.
1st Generation
At each generation, a different data subset (i.e., different window) is used.
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Fitness Evaluation
![Page 37: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/37.jpg)
Idea of Windowing in J. Bacardit et al.: Speeding-up Pittsburgh learning classifier systems: Modeling time and accuracy. PPSN 2004.
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
2nd Generation
At each generation, a different data subset (i.e., different window) is used.
Fitness Evaluation
![Page 38: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/38.jpg)
Idea of Windowing in J. Bacardit et al.: Speeding-up Pittsburgh learning classifier systems: Modeling time and accuracy. PPSN 2004.
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
3rd Generation
At each generation, a different data subset (i.e., different window) is used.
Fitness Evaluation
![Page 39: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/39.jpg)
Visual Image of WindowingA population is moving around in the training data.
Training Data = Environment
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Population
![Page 40: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/40.jpg)
Visual Image of WindowingA population is moving around in the training data.
Training Data = Environment
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Population
![Page 41: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/41.jpg)
Visual Image of WindowingA population is moving around in the training data.
Training Data = Environment
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Population
![Page 42: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/42.jpg)
Visual Image of WindowingA population is moving around in the training data.
Training Data = Environment
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Population
![Page 43: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/43.jpg)
Visual Image of WindowingA population is moving around in the training data.
Training Data = Environment
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Population
![Page 44: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/44.jpg)
Visual Image of WindowingA population is moving around in the training data.
Training Data = Environment
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
If ... Then ...If ... Then ...If ... Then ...If ... Then ...
Population
After enough evolution with a moving windowThe population does not overfit to any particular training data subset.The population may have high generalization ability.
![Page 45: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/45.jpg)
Training data
CPU
Popu
latio
n
Non-parallel Non-distributed
Our Idea: Parallel Distributed ImplementationH. Ishibuchi et al.: Parallel Distributed Hybrid Fuzzy GBML Models with Rule Set Migration and Training Data Rotation. TFS (in Press)
![Page 46: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/46.jpg)
CPU CPU CPU CPU CPU
Training data
CPU
Popu
latio
n
Non-parallel Non-distributed
Our Idea: Parallel Distributed ImplementationH. Ishibuchi et al.: Parallel Distributed Hybrid Fuzzy GBML Models with Rule Set Migration and Training Data Rotation. TFS (in Press)
(1) A population is divided into multiple subpopulations.(as in an island model)
Our Parallel Distributed Model
![Page 47: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/47.jpg)
CPU CPU CPU CPU CPU
Training data
CPU
Popu
latio
n
Non-parallel Non-distributed
Our Idea: Parallel Distributed ImplementationH. Ishibuchi et al.: Parallel Distributed Hybrid Fuzzy GBML Models with Rule Set Migration and Training Data Rotation. TFS (in Press)
Training data
(1) A population is divided into multiple subpopulations.(2) Training data are also divided into multiple subsets.
(as in the windowing method)
Our Parallel Distributed Model
![Page 48: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/48.jpg)
CPU CPU CPU CPU CPU
Training data
CPU
Popu
latio
n
Non-parallel Non-distributed
Our Idea: Parallel Distributed ImplementationH. Ishibuchi et al.: Parallel Distributed Hybrid Fuzzy GBML Models with Rule Set Migration and Training Data Rotation. TFS (in Press)
Training data
(1) A population is divided into multiple subpopulations.(2) Training data are also divided into multiple subsets.(3) An evolutionary algorithm is locally performed at each CPU.
(as in an island model)
Our Parallel Distributed Model
![Page 49: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/49.jpg)
CPU CPU CPU CPU CPU
Training data
CPU
Popu
latio
n
Our Parallel Distributed ModelNon-parallel Non-distributed
Our Idea: Parallel Distributed ImplementationH. Ishibuchi et al.: Parallel Distributed Hybrid Fuzzy GBML Models with Rule Set Migration and Training Data Rotation. TFS (in Press)
Training data
(1) A population is divided into multiple subpopulations.(2) Training data are also divided into multiple subsets.(3) An evolutionary algorithm is locally performed at each CPU.(4) Training data subsets are periodically rotated.
(e.g., every 100 generations)
![Page 50: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/50.jpg)
CPU CPU CPU CPU CPU
Training data
CPU
Popu
latio
n
Our Parallel Distributed ModelNon-parallel Non-distributed
Our Idea: Parallel Distributed ImplementationH. Ishibuchi et al.: Parallel Distributed Hybrid Fuzzy GBML Models with Rule Set Migration and Training Data Rotation. TFS (in Press)
Training data
(1) A population is divided into multiple subpopulations.(2) Training data are also divided into multiple subsets.(3) An evolutionary algorithm is locally performed at each CPU.(4) Training data subsets are periodically rotated.(5) Migration is also periodically performed.
![Page 51: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/51.jpg)
Computation Load of Four Models
EC
Fitness Evaluation
Standard Non-Parallel Model
CPU
Computation Load = EC Part + Fitness Evaluation Part
![Page 52: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/52.jpg)
Computation Load of Four Models
EC
Fitness Evaluation
Standard Non-Parallel Model
CPU
Computation Load = EC Part + Fitness Evaluation Part
EC
EC = Evolutionary Computation
= { Selection, Crossover, Mutation, Generation Update }
![Page 53: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/53.jpg)
Computation Load of Four Models
EC
Fitness Evaluation
Standard Non-Parallel Model
EC
Standard Parallel Model(Parallel Fitness Evaluation)
CPU
CPU
CPU
CPU
CPU
CPU
CPU
CPU Computation Load = EC Part
+ Fitness Evaluation Part (1/7)
![Page 54: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/54.jpg)
Computation Load of Four Models
EC
Fitness Evaluation
Standard Non-Parallel Model
EC
Standard Parallel Model(Parallel Fitness Evaluation)
EC
Windowing Model(Reduced Training Data Set)
CPU CPU
CPU
CPU
CPU
CPU
CPU
CPU
CPU
Computation Load = EC Part
+ Fitness Evaluation Part (1/7)
![Page 55: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/55.jpg)
Computation Load of Four Models
EC
Fitness Evaluation
Standard Non-Parallel Model
EC
Standard Parallel Model(Parallel Fitness Evaluation)
EC
Windowing Model(Reduced Training Data Set)
Parallel Distributed Model(Divided Population & Data Set)
CPU CPU
CPU
CPU
CPU
CPU
CPU
CPU
CPU CPU
CPU
CPU
CPU
CPU
CPU
CPU
![Page 56: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/56.jpg)
Computation Load of Four Models
EC
Fitness Evaluation
Standard Non-Parallel Model
EC
Standard Parallel Model(Parallel Fitness Evaluation)
EC
Windowing Model(Reduced Training Data Set)
Parallel Distributed Model(Divided Population & Data Set)
CPU CPU
CPU
CPU
CPU
CPU
CPU
CPU
CPU CPU
CPU
CPU
CPU
CPU
CPU
CPU
EC Part (1/7)
Fitness Evaluation Part (1/7)
![Page 57: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/57.jpg)
Contents of This Presentation
Introduction1. Basic Idea of Evolutionary Computation2. Genetics-Based Machine Learning3. Parallel Distributed Implementation4. Computation Experiments5. Conclusion
![Page 58: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/58.jpg)
Our Model in Computational Experimentswith Seven Subpopulations and Seven Data Subsets
![Page 59: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/59.jpg)
Our Model in Computational Experimentswith Seven Subpopulations and Seven Data Subsets
Seven CPUs
Seven Subpopulations of Size 30Population Size = 30 x 7 = 210
Seven Data Subsets
![Page 60: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/60.jpg)
Standard Non-Parallel Non-Distributed Modelwith a Single Population and a Single Data Set
Popu
latio
n
![Page 61: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/61.jpg)
Standard Non-Parallel Non-Distributed Modelwith a Single Population and a Single Data Set
Single CPU
Single Population of Size 210
Whole Data Set
![Page 62: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/62.jpg)
Standard Non-Parallel Non-Distributed Modelwith a Single Population and a Single Data Set
Single CPU
Single Population of Size 210
Whole Data Set
Termination Conditions: 50,000 GenerationsComputation Load: 210 x 50,000 = 10,500,000 Evaluations
(more than ten million evaluations)
![Page 63: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/63.jpg)
Comparison of Computation Load
Standard Model:Evaluation of 210 rule sets using all the training data
Parallel Distributed Model:Evaluation of 30 rule sets using one of the seven data subsets.
Computation Load on a Single CPU per Generation
![Page 64: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/64.jpg)
Comparison of Computation Load
Standard Model:Evaluation of 210 rule sets using all the training data
Parallel Distributed Model:Evaluation of 30 rule sets using one of the seven data subsets.
Computation Load on a Single CPU per Generation
1/7
1/7
![Page 65: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/65.jpg)
Comparison of Computation Load
Standard Model:Evaluation of 210 rule sets using all the training data
Parallel Distributed Model:Evaluation of 30 rule sets using one of the seven data subsets.
Computation Load on a Single CPU per Generation
1/7
1/7
Computation Load ==> 1/7 x 1/7 = 1/49 (about 2%)
![Page 66: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/66.jpg)
Data Sets in Computational ExperimentsNine Pattern Classification Problems
Name of Data Set
Number of Patterns
Number of Attributes
Number of Classes
Segment 2,310 19 7Phoneme 5,404 5 2
Page-blocks 5,472 10 5Texture 5,500 40 11
Satimage 6,435 36 6Twonorm 7,400 20 2
Ring 7,400 20 2PenBased 10,992 16 10
Magic 19,020 10 2
![Page 67: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/67.jpg)
Computation Time for 50,000 GenerationsComputation time was decreased to about 2%
Name of Data Set
StandardA minutes
Our Model B minutes
Percentage of B B/A (%)
Segment 203.66 4.69 2.30%Phoneme 439.18 13.19 3.00%
Page-blocks 204.63 4.74 2.32%Texture 766.61 15.72 2.05%
Satimage 658.89 15.38 2.33%Twonorm 856.58 7.84 0.92%
Ring 1015.04 22.52 2.22%PenBased 1520.54 35.56 2.34%
Magic 771.05 22.58 2.93%
![Page 68: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/68.jpg)
Computation Time for 50,000 GenerationsComputation time was decreased to about 2%
Name of Data Set
StandardA minutes
Our Model B minutes
Percentage of B B/A (%)
Segment 203.66 4.69 2.30%Phoneme 439.18 13.19 3.00%
Page-blocks 204.63 4.74 2.32%Texture 766.61 15.72 2.05%
Satimage 658.89 15.38 2.33%Twonorm 856.58 7.84 0.92%
Ring 1015.04 22.52 2.22%PenBased 1520.54 35.56 2.34%
Magic 771.05 22.58 2.93%
Why ?Because the population and the trainingdata were divided into seven subsets.
1/7 x 1/7 = 1/49 (about 2%)
![Page 69: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/69.jpg)
Computation Time for 50,000 GenerationsComputation time was decreased to about 2%
Name of Data Set
StandardA minutes
Our Model B minutes
Percentage of B B/A (%)
Segment 203.66 4.69 2.30%Phoneme 439.18 13.19 3.00%
Page-blocks 204.63 4.74 2.32%Texture 766.61 15.72 2.05%
Satimage 658.89 15.38 2.33%Twonorm 856.58 7.84 0.92%
Ring 1015.04 22.52 2.22%PenBased 1520.54 35.56 2.34%
Magic 771.05 22.58 2.93%
Why ?Because the population and the trainingdata were divided into seven subsets.
1/7 x 1/7 = 1/49 (about 2%)
![Page 70: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/70.jpg)
Test Data Error Rates (Results of 3x10CV)Test data accuracy was improved for six data sets
Name of Data Set
Standard(A %)
Our Model (B %)
Improvement from A: (A - B)%
Segment 5.99 5.90 0.09Phoneme 15.43 15.96 - 0.53
Page-blocks 3.81 3.62 0.19Texture 4.64 4.77 - 0.13
Satimage 15.54 12.96 2.58 Twonorm 7.36 3.39 3.97
Ring 6.73 5.25 1.48 PenBased 3.07 3.30 - 0.23
Magic 15.42 14.89 0.53
![Page 71: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/71.jpg)
Test Data Error Rates (Results of 3x10CV)Test data accuracy was improved for six data sets
Name of Data Set
Standard(A %)
Our Model (B %)
Improvement from A: (A - B)%
Segment 5.99 5.90 0.09Phoneme 15.43 15.96 - 0.53
Page-blocks 3.81 3.62 0.19Texture 4.64 4.77 - 0.13
Satimage 15.54 12.96 2.58 Twonorm 7.36 3.39 3.97
Ring 6.73 5.25 1.48 PenBased 3.07 3.30 - 0.23
Magic 15.42 14.89 0.53
![Page 72: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/72.jpg)
Q. Why did our model improve the test data accuracy ?A. Because our model improved the search ability.
Data Set Standard Our Model ImprovementSatimage 15.54% 12.96% 2.58% Tr
aini
ng D
ata
Erro
r Rat
e (%
)
Number of Generations
Our Parallel Distributed Model
Non-Parallel Non-Distributed
![Page 73: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/73.jpg)
Q. Why did our model improve the search ability ?A. Because our model maintained the diversity.
Data Set Standard Our Model ImprovementSatimage 15.54% 12.96% 2.58%
Training Data Rotation:Every 100 Generations
Rule Set Migration:Every 100 Generations
Trai
ning
Dat
a Er
ror R
ate
(%)
Number of Generations
Parallel Distributed Model
Non-Parallel Non-Distributed Model(Best = Worst: No Diversity)
The best and worst error ratesin a particular subpopulation ateach generation in a single run.
Worst
Best
![Page 74: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/74.jpg)
Effects of Rotation and Migration Intervals
Training Data Rotation: Every 100 GenerationsRule Set Migration: Every 100 Generations
![Page 75: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/75.jpg)
Effects of Rotation and Migration Intervals(Rotations in the opposite directions)
CPU
Training data rotation
Training data
Popu
latio
n
CPU CPU CPU CPU CPU CPU
Rule set migration
NoNo
![Page 76: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/76.jpg)
Effects of Rotation and Migration Intervals(Rotations in the opposite directions)
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
CPU
Training data rotation
Training data
Popu
latio
n
CPU CPU CPU CPU CPU CPU
Rule set migration
NoNo
Good Results
![Page 77: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/77.jpg)
Effects of Rotation and Migration Intervals(Rotations in the same direction)
CPU
Training data rotation
Training data
Popu
latio
n
CPU CPU CPU CPU CPU CPU
Rule set migration
NoNo
![Page 78: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/78.jpg)
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
Effects of Rotation and Migration Intervals(Rotations in the same direction)
CPU
Training data rotation
Training data
Popu
latio
n
CPU CPU CPU CPU CPU CPU
Rule set migration
NoNo
Good Results
![Page 79: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/79.jpg)
Effects of Rotation and Migration Intervals
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
Opposite Directions Same Direction
NoNo NoNo
![Page 80: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/80.jpg)
Effects of Rotation and Migration Intervals
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
Opposite Directions Same Direction
NoNo NoNo
Interaction between rotation and migration
![Page 81: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/81.jpg)
Effects of Rotation and Migration Intervals
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
500∞
200100
50500
∞
200100
50
18
16
14
12
1020
20
Opposite Directions Same Direction
Interaction between rotation and migration
NoNo NoNo
Negative Effects
![Page 82: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/82.jpg)
Contents of This Presentation
Introduction1. Basic Idea of Evolutionary Computation2. Genetics-Based Machine Learning3. Parallel Distributed Implementation4. Computation Experiments5. Conclusion
![Page 83: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/83.jpg)
Conclusion
Introduction1.We explained our parallel distributed model.2. It was shown that the computation time was decreased to 2%.3. It was shown that the test data accuracy was improved.4.We explained negative effects of the interaction between the
training data rotation and the rule set migration.
![Page 84: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/84.jpg)
Conclusion
Introduction1.We explained our parallel distributed model.2. It was shown that the computation time was decreased to 2%.3. It was shown that the test data accuracy was improved.4.We explained negative effects of the interaction between the
training data rotation and the rule set migration.
![Page 85: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/85.jpg)
Conclusion
Introduction1.We explained our parallel distributed model.2. It was shown that the computation time was decreased to 2%.3. It was shown that the test data accuracy was improved.4.We explained negative effects of the interaction between the
training data rotation and the rule set migration.
![Page 86: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/86.jpg)
Conclusion
Introduction1.We explained our parallel distributed model.2. It was shown that the computation time was decreased to 2%.3. It was shown that the test data accuracy was improved.4.We explained negative effects of the interaction between the
training data rotation and the rule set migration.
![Page 87: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/87.jpg)
Conclusion
Introduction5. A little bit different model may be also possible for learning
from locally located data bases.
Subpopulation Rotation
CPU
University A
CPU
University C
CPU
University B
CPU
University D
![Page 88: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/88.jpg)
Conclusion
Introduction6. This model can be also used for learning from changing
data bases.
Subpopulation Rotation
CPU
University A
CPU
University C
CPU
University B
CPU
University D
![Page 89: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/89.jpg)
Conclusion
Introduction
Subpopulation Rotation
CPU
University A
CPU
University C
CPU
University B
CPU
University D
6. This model can be also used for learning from changingdata bases.
![Page 90: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/90.jpg)
Conclusion
Introduction
Subpopulation Rotation
CPU
University A
CPU
University C
CPU
University B
CPU
University D
6. This model can be also used for learning from changingdata bases.
![Page 91: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/91.jpg)
Conclusion
Introduction
Subpopulation Rotation
CPU
University A
CPU
University C
CPU
University B
CPU
University D
6. This model can be also used for learning from changingdata bases.
![Page 92: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/92.jpg)
Conclusion
Introduction
Subpopulation Rotation
CPU
University A
CPU
University C
CPU
University B
CPU
University D
6. This model can be also used for learning from changingdata bases.
![Page 93: Hisao Ishibuchi: "Scalability Improvement of Genetics-Based Machine Learning to Large Data Sets"](https://reader033.fdocuments.in/reader033/viewer/2022052911/559cd3dd1a28ab4f7f8b4848/html5/thumbnails/93.jpg)
Conclusion
Thank you very much!