Downloaded 25 times
Uploaded byAl Mamun
Design cycles of pattern recognition
The document outlines the design cycles of pattern recognition, which involves a series of steps including data collection, feature selection, model selection, training, and evaluation. It discusses the importance of gathering a representative data set, choosing appropriate features based on the problem domain, selecting models based on various criteria, and the different learning paradigms. Additionally, it highlights evaluation methods to estimate classifier performance and addresses concerns like overfitting.
More Related Content
Similar to Design cycles of pattern recognition
Design cycles of pattern recognition
- 1. Design Cycles of PatternRecognition Al Mamun Khan 1
- 2. Overview Design Circle Data collection Feature Choice Model Choice Training Evaluation 2
- 3. Collect Data Choose Features ChooseModel Train Classifier Evaluate Classifier Design Circle
- 4. Collect Data Choose Features Choose Model Train Classifier Evaluate Classifier Data Collection DataCollection Collecting training and testing data How can we know when we have adequately large and representative set of samples?
- 5. Collect Data Choose Features Choose Model Train Classifier Evaluate Classifier Features Choose Features Depends on characteristics of problem domain Ideally Simple to extract Invariant to irrelevant transformation Insensitive to noise
- 6. Collect Data Choose Features Choose Model Train Classifier Evaluate Classifier Model Choose Model Domain dependence and prior information. Definition of design criteria. Parametric vs. non- parametric models. Handling of missing features. Computational complexity.
- 8. Collect Data Choose Features Choose Model Train Classifier Evaluate Classifier Training Train Classifier How can we learn the rule from data? Supervised learning Unsupervised learning Reinforcement learning: no desired category is given but the teacher provides feedback to the system such as the decision is right or wrong.
- 9. Collect Data Choose Features Choose Model Train Classifier Evaluate Classifier Evaluate Classifier EvaluateClassifier How can we estimate the performance with training samples? How can we predict the performance with future data? Problems of overfitting and generalization.