SlideShare une entreprise Scribd logo
1  sur  30
Télécharger pour lire hors ligne
Evolutionary
Computation
Research
Group



  Feature Pattern Classifier System
         Handwritten Digit Classification with LCS
      Ignas Kukenys
      Victoria University of Wellington (now University of Otago)
      Ignas@cs.otago.ac.nz

      Will N. Browne
      Victoria University of Wellington
      Will.Browne@vuw.ac.nz

      Mengjie Zhang
      Victoria University of Wellington
      Mengjie.Zhang@ecs.vuw.ac.nz
Context
l    Machine learning for Robotics:
      l    Needs to be reinforcement-based and online
      l    Preferably also adaptive and transparent
l    Learning from visual input is hard:
      l    High-dimensionality vs. sparseness of data
l    Why Learning Classifier Systems
      l    Robust reinforcement learning
      l    Limited applications for visual input

                                                         2
Goals
l    Adapt LCS to learn from image data
      l    Use image features that enable generalisation
      l    Tweak the evolutionary process
      l    Use a well known vision problem for evaluation


l    Build a classifier system for handwritten digit
      classification


                                                             3
Learning Classifier Systems
l    LCS model an agent interacting with an
      unknown environment:
      l    Agent observes a state of the environment
      l    Agent performs an action
      l    Environment provides a reward


l    The above contract constrains learning:
      l    Online: one problem instance at a time
      l    Ground truth not available (non-supervised)
                                                          4
Learning Classifier Systems




                              5
Learning Classifier Systems




                              6
Basics of LCS
l    LCS evolve a population of rules:
                   if condition(s) then action


l    Each rule also has associated properties:
      l    Predicted reward for advocated action
      l    Accuracy based on prediction error
      l    Fitness based on relative accuracy


                                                    7
Simple rule conditions
l    Traditionally LCS use 'don't care' (#) encoding:
      l    e.g. condition #1# matches states 010, 111, 110 and
            111
l    Enables rules to generalise over multiple states
l    Varying levels of generalisation:
      l    ### matches all possible states
      l    010 matches a single specific state
Naïve image classification
l    Consider binary 3x3 pixel patterns:




l    How to separate them into two classes
      based on the colour of centre point?


                                              9
Naïve image classification
l    Environment states: 9 bit messages
           l    e.g. 011100001 and 100010101




l    Two actions represent two classes: 0, 1
l    Two rules are sufficient to solve the problem:
      [### #0# ###] → 0
      [### #1# ###] → 1
                                                       10
Naïve image classification
l    Example 2: how to classify 3x3 patterns that
      have “a horizontal line of 3 white pixels”?
      [111 ### ###] → 1
      [### 111 ###] → 1
      [### ### 111] → 1
l    Example 3: how to deal with 3x3 patterns “at
      least one 0 on every row”?
        l    27 unique rules to fully describe the
              problem
                                                      11
Naïve image classification
l    Number of rules explodes for complex patterns
l    Consider 256 pixel values for grey-scale, …
l    Very limited generalisation in such conditions
l    Photographic and other “real world” images:
        l    Significantly different at “pixel level”
        l    Need more flexible conditions




                                                         12
Haar-like features




                     13
Haar-like features
l    Compute differences between pixel sums in
      rectangular regions of the image
l    Very efficient with the use of “integral image”
l    Widely used in computer vision
       l    e.g. state of the art Viola & Jones face detector
l    Can be flexibly placed at different scales and
      positions in the image
l    Enable varying levels of generalisation

                                                                 14
Haar-like feature rules
l    To obtain LCS-like rules, feature outputs need
      to be thresholded:


if (feature(type, position, scale) > threshold) then action


l    Flexible direction of comparison: < and >
l    Range: t_low < feature < t_high


                                                        15
“Messy” encoding
l    Multiple features form stronger rules:
if (feature_1 && feature_2 && feature_3 ...) then action

l    Seems to be a limit to a useful number of
      features:




                                                           16
MNIST digits dataset
l    Well known handwritten digits dataset
l    60 000 training examples, 10 classes
l    Examples from 250 subjects
l    28x28 pixel grey-scale (0..255) images
l    10 000 evaluation examples (test set, different
      subjects)



                                                    17
MNIST results




                18
MNIST results
l    Performance:
       l    Training set: 92% after 4M observations
       l    Evaluation set: 91%
l    Supervised and off-line methods reach 99%
l    Encouraging initial result for reinforcement
      learning




                                                       19
Adaptive learning




                    20
Why not 100% performance?




                            21
Improving the FPCS
l    Tournament selection
       l    Performs better than proportional RW
l    Crossover only at feature level
       l    Rules swap features, not individual attributes
l    Features start at “best” position, then mutate
       l    Instead of random position place feature where
             the output is highest
l    With all other fixes, performance still at 94%

                                                              22
Why not 100% performance?
•         Online reinforcement learning
     •      Cannot adapt rules based on known ground truth


•         Forms of complete map of all states to all
          actions to their reward, e.g. learns “not a 3”
     •      Rather than just correct state: action mapping


•         Only uses Haar-like features
     •      Could use ensemble of different features.
                                                             23
Future work
•    Inner confusion matrix to “guide” learning to
     “hard” areas of the problem
•    Test with a supervised-learning LCS,
     e.g. UCS
•    Only learn accurate positive rules, rather than
     complete mapping
•    How to deal with outliers?
•    Testing on harder image problems will likely
     reveal further challenges
                                                       24
Confusion matrix




                   25
Confusion matrix




                   26
Conclusions
•    LCS can successfully work with image data.


•    Autonomously learn the number, type, scale
     and threshold of features to use in a
     transparent manner.


•    Challenges remain to bridge the 5% gap to
     supervised learning performance

                                                  27
Demo
•    Handwritten digit classification with FPCS




                                                  28
Questions?
Basics of LCS
l    For observed state s all conditions are tested
l    Matching rules form match set [M]
l    For every action, a reward is predicted
l    An action a is chosen (random vs. best)
l    Rules in [M] advocating a form action set [A]
l    [A] is updated according to reward received
l    Rule Discovery, e.g. GA, is performed in [A] to
      evolve better rules
                                                       30

Contenu connexe

En vedette

Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...
Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...
Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...Nicola Strisciuglio
 
STAT 897D Project 2 - Final Draft
STAT 897D Project 2 - Final DraftSTAT 897D Project 2 - Final Draft
STAT 897D Project 2 - Final DraftJonathan Fivelsdal
 
Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...
Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...
Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...Spark Summit
 
Fashion Design Project
Fashion Design ProjectFashion Design Project
Fashion Design ProjectOlivia Rumao
 
The literature review
The literature reviewThe literature review
The literature reviewBarryCRNA
 
Amazings 4 - Octubre Noviembre Diciembre - 2011
Amazings 4 - Octubre Noviembre Diciembre - 2011Amazings 4 - Octubre Noviembre Diciembre - 2011
Amazings 4 - Octubre Noviembre Diciembre - 2011degarden
 
Sub Saharan Africa Vocab GEN
Sub Saharan Africa Vocab GENSub Saharan Africa Vocab GEN
Sub Saharan Africa Vocab GENordovensky
 
F I G E S Company Presentation (21
F I G E S  Company  Presentation (21F I G E S  Company  Presentation (21
F I G E S Company Presentation (21akifgorur
 
IELA- International Exhibition Logistics Association
IELA- International Exhibition Logistics AssociationIELA- International Exhibition Logistics Association
IELA- International Exhibition Logistics AssociationLausanne Montreux Congress
 
Ceran Sherborne
Ceran SherborneCeran Sherborne
Ceran SherborneASTEX
 
Top 8 it support engineer resume samples
Top 8 it support engineer resume samplesTop 8 it support engineer resume samples
Top 8 it support engineer resume samplesjomcoret
 
Stamp Auction to be Held on 17th Aug’14 in the UK - TonyLester
Stamp Auction to be Held on 17th Aug’14 in the UK - TonyLesterStamp Auction to be Held on 17th Aug’14 in the UK - TonyLester
Stamp Auction to be Held on 17th Aug’14 in the UK - TonyLesterTony Lester
 

En vedette (20)

Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...
Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...
Cascade classifiers trained on gammatonegrams for reliably detecting audio ev...
 
STAT 897D Project 2 - Final Draft
STAT 897D Project 2 - Final DraftSTAT 897D Project 2 - Final Draft
STAT 897D Project 2 - Final Draft
 
Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...
Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...
Building a Location Based Social Graph in Spark at InMobi-(Seinjuti Chatterje...
 
Ada boost
Ada boostAda boost
Ada boost
 
Fashion Design Project
Fashion Design ProjectFashion Design Project
Fashion Design Project
 
Machine Learning
Machine LearningMachine Learning
Machine Learning
 
The literature review
The literature reviewThe literature review
The literature review
 
Decision tree
Decision treeDecision tree
Decision tree
 
Amazings 4 - Octubre Noviembre Diciembre - 2011
Amazings 4 - Octubre Noviembre Diciembre - 2011Amazings 4 - Octubre Noviembre Diciembre - 2011
Amazings 4 - Octubre Noviembre Diciembre - 2011
 
Sub Saharan Africa Vocab GEN
Sub Saharan Africa Vocab GENSub Saharan Africa Vocab GEN
Sub Saharan Africa Vocab GEN
 
Zhang Wenyu, Work Samples
Zhang Wenyu, Work SamplesZhang Wenyu, Work Samples
Zhang Wenyu, Work Samples
 
Communiqué de Presse: BILAN DU MIDEST 2014
Communiqué de Presse: BILAN DU MIDEST 2014Communiqué de Presse: BILAN DU MIDEST 2014
Communiqué de Presse: BILAN DU MIDEST 2014
 
F I G E S Company Presentation (21
F I G E S  Company  Presentation (21F I G E S  Company  Presentation (21
F I G E S Company Presentation (21
 
IELA- International Exhibition Logistics Association
IELA- International Exhibition Logistics AssociationIELA- International Exhibition Logistics Association
IELA- International Exhibition Logistics Association
 
Content Marketing
Content MarketingContent Marketing
Content Marketing
 
Pi fs 2012 2013
Pi fs 2012 2013Pi fs 2012 2013
Pi fs 2012 2013
 
Ceran Sherborne
Ceran SherborneCeran Sherborne
Ceran Sherborne
 
Top 8 it support engineer resume samples
Top 8 it support engineer resume samplesTop 8 it support engineer resume samples
Top 8 it support engineer resume samples
 
Stamp Auction to be Held on 17th Aug’14 in the UK - TonyLester
Stamp Auction to be Held on 17th Aug’14 in the UK - TonyLesterStamp Auction to be Held on 17th Aug’14 in the UK - TonyLester
Stamp Auction to be Held on 17th Aug’14 in the UK - TonyLester
 
Bc hc
Bc hcBc hc
Bc hc
 

Similaire à Confusion Matrices for Improving Performance of Feature Pattern Classifier Systems

MLIP - Chapter 5 - Detection, Segmentation, Captioning
MLIP - Chapter 5 - Detection, Segmentation, CaptioningMLIP - Chapter 5 - Detection, Segmentation, Captioning
MLIP - Chapter 5 - Detection, Segmentation, CaptioningCharles Deledalle
 
Artificial Intelligence, Machine Learning and Deep Learning
Artificial Intelligence, Machine Learning and Deep LearningArtificial Intelligence, Machine Learning and Deep Learning
Artificial Intelligence, Machine Learning and Deep LearningSujit Pal
 
Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017
Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017
Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017MLconf
 
General Tips for participating Kaggle Competitions
General Tips for participating Kaggle CompetitionsGeneral Tips for participating Kaggle Competitions
General Tips for participating Kaggle CompetitionsMark Peng
 
Intro to Deep Reinforcement Learning
Intro to Deep Reinforcement LearningIntro to Deep Reinforcement Learning
Intro to Deep Reinforcement LearningKhaled Saleh
 
Convolutional neural networks 이론과 응용
Convolutional neural networks 이론과 응용Convolutional neural networks 이론과 응용
Convolutional neural networks 이론과 응용홍배 김
 
Graph Analysis Beyond Linear Algebra
Graph Analysis Beyond Linear AlgebraGraph Analysis Beyond Linear Algebra
Graph Analysis Beyond Linear AlgebraJason Riedy
 
Two strategies for large-scale multi-label classification on the YouTube-8M d...
Two strategies for large-scale multi-label classification on the YouTube-8M d...Two strategies for large-scale multi-label classification on the YouTube-8M d...
Two strategies for large-scale multi-label classification on the YouTube-8M d...Dalei Li
 
Fundamental of deep learning
Fundamental of deep learningFundamental of deep learning
Fundamental of deep learningStanley Wang
 
Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...Lu Jiang
 
Machine Learning Essentials Demystified part1 | Big Data Demystified
Machine Learning Essentials Demystified part1 | Big Data DemystifiedMachine Learning Essentials Demystified part1 | Big Data Demystified
Machine Learning Essentials Demystified part1 | Big Data DemystifiedOmid Vahdaty
 
物件偵測與辨識技術
物件偵測與辨識技術物件偵測與辨識技術
物件偵測與辨識技術CHENHuiMei
 
Paper overview: "Deep Residual Learning for Image Recognition"
Paper overview: "Deep Residual Learning for Image Recognition"Paper overview: "Deep Residual Learning for Image Recognition"
Paper overview: "Deep Residual Learning for Image Recognition"Ilya Kuzovkin
 
Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)
Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)
Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)Anmol Dwivedi
 

Similaire à Confusion Matrices for Improving Performance of Feature Pattern Classifier Systems (20)

Supervised Learning.pptx
Supervised Learning.pptxSupervised Learning.pptx
Supervised Learning.pptx
 
MLIP - Chapter 5 - Detection, Segmentation, Captioning
MLIP - Chapter 5 - Detection, Segmentation, CaptioningMLIP - Chapter 5 - Detection, Segmentation, Captioning
MLIP - Chapter 5 - Detection, Segmentation, Captioning
 
Artificial Intelligence, Machine Learning and Deep Learning
Artificial Intelligence, Machine Learning and Deep LearningArtificial Intelligence, Machine Learning and Deep Learning
Artificial Intelligence, Machine Learning and Deep Learning
 
Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017
Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017
Matineh Shaker, Artificial Intelligence Scientist, Bonsai at MLconf SF 2017
 
General Tips for participating Kaggle Competitions
General Tips for participating Kaggle CompetitionsGeneral Tips for participating Kaggle Competitions
General Tips for participating Kaggle Competitions
 
Intro to Deep Reinforcement Learning
Intro to Deep Reinforcement LearningIntro to Deep Reinforcement Learning
Intro to Deep Reinforcement Learning
 
K Nearest Neighbor Algorithm
K Nearest Neighbor AlgorithmK Nearest Neighbor Algorithm
K Nearest Neighbor Algorithm
 
Convolutional neural networks 이론과 응용
Convolutional neural networks 이론과 응용Convolutional neural networks 이론과 응용
Convolutional neural networks 이론과 응용
 
Graph Analysis Beyond Linear Algebra
Graph Analysis Beyond Linear AlgebraGraph Analysis Beyond Linear Algebra
Graph Analysis Beyond Linear Algebra
 
Two strategies for large-scale multi-label classification on the YouTube-8M d...
Two strategies for large-scale multi-label classification on the YouTube-8M d...Two strategies for large-scale multi-label classification on the YouTube-8M d...
Two strategies for large-scale multi-label classification on the YouTube-8M d...
 
Fundamental of deep learning
Fundamental of deep learningFundamental of deep learning
Fundamental of deep learning
 
Computer Engineer Master Project
Computer Engineer Master ProjectComputer Engineer Master Project
Computer Engineer Master Project
 
Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...
 
lec6a.ppt
lec6a.pptlec6a.ppt
lec6a.ppt
 
Machine Learning Essentials Demystified part1 | Big Data Demystified
Machine Learning Essentials Demystified part1 | Big Data DemystifiedMachine Learning Essentials Demystified part1 | Big Data Demystified
Machine Learning Essentials Demystified part1 | Big Data Demystified
 
lec26.pptx
lec26.pptxlec26.pptx
lec26.pptx
 
物件偵測與辨識技術
物件偵測與辨識技術物件偵測與辨識技術
物件偵測與辨識技術
 
Paper overview: "Deep Residual Learning for Image Recognition"
Paper overview: "Deep Residual Learning for Image Recognition"Paper overview: "Deep Residual Learning for Image Recognition"
Paper overview: "Deep Residual Learning for Image Recognition"
 
convolutional_rbm.ppt
convolutional_rbm.pptconvolutional_rbm.ppt
convolutional_rbm.ppt
 
Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)
Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)
Inference & Learning in Linear-Chain Conditional Random Fields (CRFs)
 

Plus de Daniele Loiacono

GPUs for GEC Competition @ GECCO-2013
GPUs for GEC Competition @ GECCO-2013GPUs for GEC Competition @ GECCO-2013
GPUs for GEC Competition @ GECCO-2013Daniele Loiacono
 
EvoRobocode Competition @ GECCO-2013
EvoRobocode Competition @ GECCO-2013EvoRobocode Competition @ GECCO-2013
EvoRobocode Competition @ GECCO-2013Daniele Loiacono
 
2013 Simulated Car Racing @ GECCO-2013
2013 Simulated Car Racing @ GECCO-20132013 Simulated Car Racing @ GECCO-2013
2013 Simulated Car Racing @ GECCO-2013Daniele Loiacono
 
2012 Simulated Car Racing Championship @ CIG-2012
2012 Simulated Car Racing Championship @ CIG-20122012 Simulated Car Racing Championship @ CIG-2012
2012 Simulated Car Racing Championship @ CIG-2012Daniele Loiacono
 
2012 Simulated Car Racing Championship @ GECCO-2012
2012 Simulated Car Racing Championship @ GECCO-20122012 Simulated Car Racing Championship @ GECCO-2012
2012 Simulated Car Racing Championship @ GECCO-2012Daniele Loiacono
 
2012 Simulated Car Racing Championship @ Evo*-2012
2012 Simulated Car Racing Championship @ Evo*-20122012 Simulated Car Racing Championship @ Evo*-2012
2012 Simulated Car Racing Championship @ Evo*-2012Daniele Loiacono
 
Computational Intelligence in Games Tutorial @GECCO2012
Computational Intelligence in Games Tutorial @GECCO2012Computational Intelligence in Games Tutorial @GECCO2012
Computational Intelligence in Games Tutorial @GECCO2012Daniele Loiacono
 
XCSF with Local Deletion: Preventing Detrimental Forgetting
XCSF with Local Deletion: Preventing Detrimental ForgettingXCSF with Local Deletion: Preventing Detrimental Forgetting
XCSF with Local Deletion: Preventing Detrimental ForgettingDaniele Loiacono
 
Testing learning classifier systems
Testing learning classifier systemsTesting learning classifier systems
Testing learning classifier systemsDaniele Loiacono
 
Random Artificial Incorporation of Noise in a Learning Classifier System Envi...
Random Artificial Incorporation of Noise in a Learning Classifier System Envi...Random Artificial Incorporation of Noise in a Learning Classifier System Envi...
Random Artificial Incorporation of Noise in a Learning Classifier System Envi...Daniele Loiacono
 
Introducing LCS to Digital Design Verification
Introducing LCS to Digital Design VerificationIntroducing LCS to Digital Design Verification
Introducing LCS to Digital Design VerificationDaniele Loiacono
 
A temporal classifier system using spiking neural networks
A temporal classifier system using spiking neural networksA temporal classifier system using spiking neural networks
A temporal classifier system using spiking neural networksDaniele Loiacono
 
Automatically Defined Functions for Learning Classifier Systems
Automatically Defined Functions for Learning Classifier SystemsAutomatically Defined Functions for Learning Classifier Systems
Automatically Defined Functions for Learning Classifier SystemsDaniele Loiacono
 
Voting Based Learning Classifier System for Multi-Label Classification
Voting Based Learning Classifier System for Multi-Label ClassificationVoting Based Learning Classifier System for Multi-Label Classification
Voting Based Learning Classifier System for Multi-Label ClassificationDaniele Loiacono
 
2011 Simulated Car Racing Championship @ GECCO-2011
2011 Simulated Car Racing Championship @ GECCO-20112011 Simulated Car Racing Championship @ GECCO-2011
2011 Simulated Car Racing Championship @ GECCO-2011Daniele Loiacono
 
2010 Simulated Car Racing Championship @ CIG-2010
2010 Simulated Car Racing Championship @ CIG-20102010 Simulated Car Racing Championship @ CIG-2010
2010 Simulated Car Racing Championship @ CIG-2010Daniele Loiacono
 
2010 Simulated Car Racing Championship @ GECCO-2010
2010 Simulated Car Racing Championship @ GECCO-20102010 Simulated Car Racing Championship @ GECCO-2010
2010 Simulated Car Racing Championship @ GECCO-2010Daniele Loiacono
 
2010 Simulated Car Racing Championship @ WCCI-2010
2010 Simulated Car Racing Championship @ WCCI-20102010 Simulated Car Racing Championship @ WCCI-2010
2010 Simulated Car Racing Championship @ WCCI-2010Daniele Loiacono
 
Car Setup Optimization Competition @ EvoStar 2010
Car Setup Optimization Competition @ EvoStar 2010Car Setup Optimization Competition @ EvoStar 2010
Car Setup Optimization Competition @ EvoStar 2010Daniele Loiacono
 

Plus de Daniele Loiacono (20)

GPUs for GEC Competition @ GECCO-2013
GPUs for GEC Competition @ GECCO-2013GPUs for GEC Competition @ GECCO-2013
GPUs for GEC Competition @ GECCO-2013
 
EvoRobocode Competition @ GECCO-2013
EvoRobocode Competition @ GECCO-2013EvoRobocode Competition @ GECCO-2013
EvoRobocode Competition @ GECCO-2013
 
2013 Simulated Car Racing @ GECCO-2013
2013 Simulated Car Racing @ GECCO-20132013 Simulated Car Racing @ GECCO-2013
2013 Simulated Car Racing @ GECCO-2013
 
2012 Simulated Car Racing Championship @ CIG-2012
2012 Simulated Car Racing Championship @ CIG-20122012 Simulated Car Racing Championship @ CIG-2012
2012 Simulated Car Racing Championship @ CIG-2012
 
2012 Simulated Car Racing Championship @ GECCO-2012
2012 Simulated Car Racing Championship @ GECCO-20122012 Simulated Car Racing Championship @ GECCO-2012
2012 Simulated Car Racing Championship @ GECCO-2012
 
2012 Simulated Car Racing Championship @ Evo*-2012
2012 Simulated Car Racing Championship @ Evo*-20122012 Simulated Car Racing Championship @ Evo*-2012
2012 Simulated Car Racing Championship @ Evo*-2012
 
Computational Intelligence in Games Tutorial @GECCO2012
Computational Intelligence in Games Tutorial @GECCO2012Computational Intelligence in Games Tutorial @GECCO2012
Computational Intelligence in Games Tutorial @GECCO2012
 
XCSF with Local Deletion: Preventing Detrimental Forgetting
XCSF with Local Deletion: Preventing Detrimental ForgettingXCSF with Local Deletion: Preventing Detrimental Forgetting
XCSF with Local Deletion: Preventing Detrimental Forgetting
 
Testing learning classifier systems
Testing learning classifier systemsTesting learning classifier systems
Testing learning classifier systems
 
Random Artificial Incorporation of Noise in a Learning Classifier System Envi...
Random Artificial Incorporation of Noise in a Learning Classifier System Envi...Random Artificial Incorporation of Noise in a Learning Classifier System Envi...
Random Artificial Incorporation of Noise in a Learning Classifier System Envi...
 
One Step Fits All
One Step Fits AllOne Step Fits All
One Step Fits All
 
Introducing LCS to Digital Design Verification
Introducing LCS to Digital Design VerificationIntroducing LCS to Digital Design Verification
Introducing LCS to Digital Design Verification
 
A temporal classifier system using spiking neural networks
A temporal classifier system using spiking neural networksA temporal classifier system using spiking neural networks
A temporal classifier system using spiking neural networks
 
Automatically Defined Functions for Learning Classifier Systems
Automatically Defined Functions for Learning Classifier SystemsAutomatically Defined Functions for Learning Classifier Systems
Automatically Defined Functions for Learning Classifier Systems
 
Voting Based Learning Classifier System for Multi-Label Classification
Voting Based Learning Classifier System for Multi-Label ClassificationVoting Based Learning Classifier System for Multi-Label Classification
Voting Based Learning Classifier System for Multi-Label Classification
 
2011 Simulated Car Racing Championship @ GECCO-2011
2011 Simulated Car Racing Championship @ GECCO-20112011 Simulated Car Racing Championship @ GECCO-2011
2011 Simulated Car Racing Championship @ GECCO-2011
 
2010 Simulated Car Racing Championship @ CIG-2010
2010 Simulated Car Racing Championship @ CIG-20102010 Simulated Car Racing Championship @ CIG-2010
2010 Simulated Car Racing Championship @ CIG-2010
 
2010 Simulated Car Racing Championship @ GECCO-2010
2010 Simulated Car Racing Championship @ GECCO-20102010 Simulated Car Racing Championship @ GECCO-2010
2010 Simulated Car Racing Championship @ GECCO-2010
 
2010 Simulated Car Racing Championship @ WCCI-2010
2010 Simulated Car Racing Championship @ WCCI-20102010 Simulated Car Racing Championship @ WCCI-2010
2010 Simulated Car Racing Championship @ WCCI-2010
 
Car Setup Optimization Competition @ EvoStar 2010
Car Setup Optimization Competition @ EvoStar 2010Car Setup Optimization Competition @ EvoStar 2010
Car Setup Optimization Competition @ EvoStar 2010
 

Dernier

"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek SchlawackFwdays
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationRidwan Fadjar
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 3652toLead Limited
 
Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Commit University
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationSlibray Presentation
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Patryk Bandurski
 
DevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsDevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsSergiu Bodiu
 
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024BookNet Canada
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfAddepto
 
Search Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdfSearch Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdfRankYa
 
SAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptxSAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptxNavinnSomaal
 
CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):comworks
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsRizwan Syed
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationSafe Software
 
Dev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio WebDev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio WebUiPathCommunity
 
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage CostLeverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage CostZilliz
 
Vertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering TipsVertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering TipsMiki Katsuragi
 
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks..."LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...Fwdays
 
Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Enterprise Knowledge
 
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr LapshynFwdays
 

Dernier (20)

"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 Presentation
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365
 
Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck Presentation
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
 
DevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsDevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platforms
 
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdf
 
Search Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdfSearch Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdf
 
SAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptxSAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptx
 
CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL Certs
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
 
Dev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio WebDev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio Web
 
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage CostLeverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
 
Vertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering TipsVertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering Tips
 
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks..."LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
 
Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024
 
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
 

Confusion Matrices for Improving Performance of Feature Pattern Classifier Systems

  • 1. Evolutionary Computation Research Group Feature Pattern Classifier System Handwritten Digit Classification with LCS Ignas Kukenys Victoria University of Wellington (now University of Otago) Ignas@cs.otago.ac.nz Will N. Browne Victoria University of Wellington Will.Browne@vuw.ac.nz Mengjie Zhang Victoria University of Wellington Mengjie.Zhang@ecs.vuw.ac.nz
  • 2. Context l  Machine learning for Robotics: l  Needs to be reinforcement-based and online l  Preferably also adaptive and transparent l  Learning from visual input is hard: l  High-dimensionality vs. sparseness of data l  Why Learning Classifier Systems l  Robust reinforcement learning l  Limited applications for visual input 2
  • 3. Goals l  Adapt LCS to learn from image data l  Use image features that enable generalisation l  Tweak the evolutionary process l  Use a well known vision problem for evaluation l  Build a classifier system for handwritten digit classification 3
  • 4. Learning Classifier Systems l  LCS model an agent interacting with an unknown environment: l  Agent observes a state of the environment l  Agent performs an action l  Environment provides a reward l  The above contract constrains learning: l  Online: one problem instance at a time l  Ground truth not available (non-supervised) 4
  • 7. Basics of LCS l  LCS evolve a population of rules: if condition(s) then action l  Each rule also has associated properties: l  Predicted reward for advocated action l  Accuracy based on prediction error l  Fitness based on relative accuracy 7
  • 8. Simple rule conditions l  Traditionally LCS use 'don't care' (#) encoding: l  e.g. condition #1# matches states 010, 111, 110 and 111 l  Enables rules to generalise over multiple states l  Varying levels of generalisation: l  ### matches all possible states l  010 matches a single specific state
  • 9. Naïve image classification l  Consider binary 3x3 pixel patterns: l  How to separate them into two classes based on the colour of centre point? 9
  • 10. Naïve image classification l  Environment states: 9 bit messages l  e.g. 011100001 and 100010101 l  Two actions represent two classes: 0, 1 l  Two rules are sufficient to solve the problem: [### #0# ###] → 0 [### #1# ###] → 1 10
  • 11. Naïve image classification l  Example 2: how to classify 3x3 patterns that have “a horizontal line of 3 white pixels”? [111 ### ###] → 1 [### 111 ###] → 1 [### ### 111] → 1 l  Example 3: how to deal with 3x3 patterns “at least one 0 on every row”? l  27 unique rules to fully describe the problem 11
  • 12. Naïve image classification l  Number of rules explodes for complex patterns l  Consider 256 pixel values for grey-scale, … l  Very limited generalisation in such conditions l  Photographic and other “real world” images: l  Significantly different at “pixel level” l  Need more flexible conditions 12
  • 14. Haar-like features l  Compute differences between pixel sums in rectangular regions of the image l  Very efficient with the use of “integral image” l  Widely used in computer vision l  e.g. state of the art Viola & Jones face detector l  Can be flexibly placed at different scales and positions in the image l  Enable varying levels of generalisation 14
  • 15. Haar-like feature rules l  To obtain LCS-like rules, feature outputs need to be thresholded: if (feature(type, position, scale) > threshold) then action l  Flexible direction of comparison: < and > l  Range: t_low < feature < t_high 15
  • 16. “Messy” encoding l  Multiple features form stronger rules: if (feature_1 && feature_2 && feature_3 ...) then action l  Seems to be a limit to a useful number of features: 16
  • 17. MNIST digits dataset l  Well known handwritten digits dataset l  60 000 training examples, 10 classes l  Examples from 250 subjects l  28x28 pixel grey-scale (0..255) images l  10 000 evaluation examples (test set, different subjects) 17
  • 19. MNIST results l  Performance: l  Training set: 92% after 4M observations l  Evaluation set: 91% l  Supervised and off-line methods reach 99% l  Encouraging initial result for reinforcement learning 19
  • 21. Why not 100% performance? 21
  • 22. Improving the FPCS l  Tournament selection l  Performs better than proportional RW l  Crossover only at feature level l  Rules swap features, not individual attributes l  Features start at “best” position, then mutate l  Instead of random position place feature where the output is highest l  With all other fixes, performance still at 94% 22
  • 23. Why not 100% performance? •  Online reinforcement learning •  Cannot adapt rules based on known ground truth •  Forms of complete map of all states to all actions to their reward, e.g. learns “not a 3” •  Rather than just correct state: action mapping •  Only uses Haar-like features •  Could use ensemble of different features. 23
  • 24. Future work •  Inner confusion matrix to “guide” learning to “hard” areas of the problem •  Test with a supervised-learning LCS, e.g. UCS •  Only learn accurate positive rules, rather than complete mapping •  How to deal with outliers? •  Testing on harder image problems will likely reveal further challenges 24
  • 27. Conclusions •  LCS can successfully work with image data. •  Autonomously learn the number, type, scale and threshold of features to use in a transparent manner. •  Challenges remain to bridge the 5% gap to supervised learning performance 27
  • 28. Demo •  Handwritten digit classification with FPCS 28
  • 30. Basics of LCS l  For observed state s all conditions are tested l  Matching rules form match set [M] l  For every action, a reward is predicted l  An action a is chosen (random vs. best) l  Rules in [M] advocating a form action set [A] l  [A] is updated according to reward received l  Rule Discovery, e.g. GA, is performed in [A] to evolve better rules 30