Tentative syllabus for the subject, “Machine Learning for Biological Systems (MLBS)”
1. Foundation of Machine learning: Turing Machine, Concepts of John von Neumann, computation of
amount of learning of a machine, common biological systems as platform for implementation.
2. Concept of supervised and unsupervised learning, concept of clusters and classes, concept of training
and testing.
3. Statistical Machine Learning:
i) Design of rule based expert system, knowledge engineering, forward chaining and backward
chaining inference techniques.
ii) Application of rule based system to discover knowledge from data, concept of clustering
(hierarchical and partition based), condition to find best clusters. Various clustering techniques
(aggomerative hierarchical, k-means clustering)
iii) Important components of a classifier, Probabilistic classifier, Bayesian classifier, Nearest
Neighbor Classifier, Discriminant Function Analysis (Linear and non-linear) as precursor to
Artificial Neural Network
iv) Hidden Markov Models and applications
v) Applications to i) discriminate Exon from Intron, ii) to predict secondary structures of
proteins, iii) discover group of genes having similar up-regulation or down-regulation pattern
from micro-array data.
4. Soft computing method based machine learning
i) Artificial Neural Network for clustering and classification, local optimization of ANN weights,
Self Organized Map (SOM), Back propagation network, Hopfield network.
ii) Genetic algorithm for optimizing parameters of classifiers.
iii) Support Vector Machine foundation, constrained local optimization using Lagrange
Multiplier, application on 2 and more than 2 classes.
iv) Over fitting and Cross validation
v) Application on the same items given in 3.v, as described above.
References:
i) “Pattern recognition and image analysis” by Earl Gose.
ii)”Pattern Classification” by Duda, Richard and David Stork
iii)”Machine Learning” by Mitchell and Tom
iv) “Artificial Intelligence and Molecular Biology”, Edited by Lawrence E. Hunter, 1993, MIT Press