Pattern Classification All materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John.
Published byModified over 4 years ago
Presentation on theme: "Pattern Classification All materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John."— Presentation transcript:
1 Pattern Classification All materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John Wiley & Sons, with the permission of the authors and the publisher
2 Intro to Pattern Recognition Chapter 1: Sections 1.1-1.6 Machine PerceptionAn ExamplePattern Recognition SystemsThe Design CycleLearning and AdaptationConclusion
3 Machine Perception Build a machine that can recognize patterns: Speech recognitionFingerprint identificationOCR (Optical Character Recognition)DNA sequence identification
4 An Example “Sorting incoming Fish on a conveyor, according to species using optical sensing”Sea bassSpeciesSalmon
6 Problem Analysis Take sample images … to extract features LengthLightnessWidthNumber and shape of finsPosition of the mouthetc…= set of features to use in our classifier!
7 Preprocessing Use a segmentation operation to isolate fish from one anotherand from the backgroundInformation from each single fish is sent to a feature extractor, whose purpose is to reduce the data by measuring certain featuresThe features are passed to a classifier
20 Pattern Recognition Systems SensingUsing transducer (camera, microphone, …)PR system depends of the bandwidth, the resolution sensitivity distortion of the transducerSegmentation and groupingPatterns should be well separated (should not overlap)
22 Feature extraction Classification Post Processing Discriminative featuresWant features INVARIANT wrt translation, rotation, scale.ClassificationUsing feature vector (provided by feature extractor) to assign given object to a categoryPost ProcessingExploit context info not from the target pattern itselfto improve performance
23 The Design Cycle Data collection Feature Choice Model Choice Training EvaluationComputational Complexity
25 Data CollectionHow do we know when we have collected an adequately large and representative set of examples for training and testing the system?
26 Feature Choice Depends on characteristics of problem domain. Ideally… Simple to extractInvariant to irrelevant transformationInsensitive to noise.
27 Model Choice If not satisfied with performance (EG, of our fish classifier)Consider another class of model
28 Training Use data to obtain good classifier identify best modeldetermine appropriate parametersMany procedures for training classifiers and choosing models
29 Evaluation Measure error rate ~= performanceMay suggest switching from one set of features to another one
30 Computational Complexity Trade-off between computational ease and performance?How algorithm scales as function of number of features, patterns or categories?
31 Learning and Adaptation Supervised learningA teacher provides a category label or cost for each pattern in the training setUnsupervised learningThe system forms clusters or “natural groupings” of the input patterns
32 ConclusionReader seems to be overwhelmed by the number, complexity and magnitude of the sub-problems of Pattern RecognitionMany of these sub-problems can indeed be solvedMany fascinating unsolved problems still remain