Presentation on theme: "Rerun of machine learning Clustering and pattern recognition."— Presentation transcript:
Rerun of machine learning Clustering and pattern recognition
Wikipedia entry on machine learning 7.1 Decision tree learning 7.2 Association rule learning 7.3 Artificial neural networks 7.4 Genetic programming 7.5 Inductive logic programming 7.6 Support vector machines 7.7 Clustering 7.8 Bayesian networks 7.9 Reinforcement learning 7.10 Representation learning 7.11 Sparse Dictionary Learning And many are still missing (ant colonies; game theory; Laplace approximations; maximum entropy
Supervised versus unsupervised Some methods really learn by themselves, but others are based on a training and testing set.
Clustering or rules Most times machine learning is used to cluster groups of data. It is also possible to find trends or rules with no clusters involved.
Force fields There is a whole class of machine learning algorithms that use force fields (especially if clustering is involved). These will be discussed separately.
Difficulty It normally is not easy to determine which method to use given the problem at hand. Some rules of thumb: 1) If you feel that a force field could do the job, but you also feel that there is a non-linear relation between data and response, then use an artificial neural network. 2) When dealing with optimisation in a high dimensional space, look into genetic algorithms.
3) When clustering things without clearly detectable clusters, look into SVMs. 4) When you have much more data than is needed, look into random forest methods. 5) When it seems clear how to deal with the data, but there are too many choices, look into decision tree methods. 6) When dealing with sequence(-like