ISLab Intelligent Systems Lab Piet van Remortel piet.vanremortel@ua.ac.be 1
Overview Who we are What we do Applied Machine learning Bio-informatics What we look for 2
Who ? Prof. Dr. Alain Verschoren Who we are Dr. Piet van Remortel What we do Applied Koenraad Van Leemput ML Tim Van den Bulcke (with ESAT-KUL) Bio-i Dr. Elmer Fernandez (Arg.) What we look for 3
What we do Who we are (Applied) machine learning What we do History in evolutionary computation Applied ML More recent: ML applications such as bio-informatics and predictive Bio-i toxicology What we look for 4
Machine Learning Wikipedia says Who we are “a broad subfield of artificial intelligence” What we do “concerned with the development of algorithms and techniques” Machine Learning “which allow computers to learn” Bio-i applications search engines, bioinformatics, What we detecting credit card fraud, stock market look for analysis, classifying DNA sequences, speech and handwriting recognition, ... 5
Machine Learning What is it good for? Who we are extract knowledge from given data What we do typically “hard” problems computationally hard Machine Learning enumeration of possibilities takes forever etc lack of traditional formalisms Bio-i traditional statistics What we complete mathematical models of the problem look for high dimensions, discontinuities, ... 6
Genetic Algorithms Who we are Use biological evolution as a metaphor for search/ optimization What we do ‘evolve’ solutions to problems Machine is an existing biological organism a ‘solution’ to Learning the problem of surviving in the environment ? iterate Bio-i generate diversity What we select for some fitness measure look for 7
Genetic Algorithms Who we are What we do Machine Learning Bio-i What we look for 8
Genetic Algorithms PRO very robust : solve a problem given some Who we are fitness measure What we do actually means: can cope with non-convex fitness landscapes Machine actually means: can solve problems where Learning good solutions are not necessarily alike each other Bio-i actully means: solving the problem is What we harder then ‘climbing one hill’ in the look for landscape can be easily parallellized (CalcUA !) 9
Genetic Algorithms Who we are What we do CON inherently slow Machine probabilistic in nature Learning true mechanics not well understood Bio-i ‘it just works’ (more or less) What we look for 10
Genetic Algorithms What we used to do with GAs Who we are study fitness landscapes What we do predict problem hardness etc What we do now Machine Learning apply ! e.g. Feature selection in predictive Bio-i toxicology use a number of genes to characterise an What we look for unknown toxic compound 11
Genetic Algorithms Typical applications of GAs: Who we are high dimensional problem What we do lots of interaction between the different elements of the solution Machine the impact of one parameter depends on the Learning value of other parameters Bio-i e.g. hot-water tap in the shower possibility to assign numerical quality to a What we candidate solution look for 12
Classification A Who we are What we do B Machine Learning Bio-i C What we look for 13 13
Classification A Who we are What we do ? B Machine Learning Bio-i C What we look for 13 13
Classification A Who we are What we do ? B Machine Learning correct Bio-i C features What we look for 13 13
Classification A Who we are What we do ? B Machine Learning Bio-i C What we look for 13 13
Classification A Who we are What we do ? B Machine Learning domain Bio-i C knowledge What we look for 13 13
Classification A Who we are What we do ? B Machine Learning Bio-i C What we look for 13 13
Classification A Who we are What we do ? B Machine Learning characterisation Bio-i C (multiple classes) What we look for 13 13
Classification of toxic compounds Cooperation with EB&T (De Coen/Blust) Who we are Goal: Characterise unknown toxic compound by means of genetic What we do expression fingerprint(s) Machine Learning 40-50 known chemicals as training set Bio-i goal: pre-compliance screening for toxic mode of action What we look for algorithmic foundation: SVM/PLS/GP 14 14
Classifiers Many classifiers around Who we are prototype of two promosing techniques What we do SVM : support vector machine Machine Learning PLS : partial least squares Bio-i more to be added Genetic programming What we look for ... 15 15
PLS: Partial least squares regression technique relate input and output matrix Who we are origin: chemistry (spectral analysis) What we do pro Machine handles collinear components in X Learning fast Bio-i What we Y X look for 16 16
SVM: Support Vector Machines Who we are What we do Machine Learning Bio-i What we look for 17 17
SVM: Support Vector Machines Who we are Classification technique What we do Machine Learning Bio-i What we look for 17 17
SVM: Support Vector Machines Who we are Classification technique What we do Mathematical foundation Machine Learning Bio-i What we look for 17 17
SVM: Support Vector Machines Who we are Classification technique What we do Mathematical foundation Machine origin: Statistical Learning Theory Learning Bio-i What we look for 17 17
SVM: Support Vector Machines Who we are Classification technique What we do Mathematical foundation Machine origin: Statistical Learning Theory Learning Bio-i bottom line: linear separability by projection to high-dimensional space What we look for 17 17
Linear separability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 18 18
Linear separability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 18 18
Linear separability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 18 18
Linear separability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 18 18
Linear separability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 18 18
Linear separability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 18 18
Linear inseparability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 19 19
Linear inseparability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 19 19
Linear inseparability (in 2D) Who we are What we do ??? Machine Learning Bio-i What we look for 19 19
Linear inseparability (in 2D) Who we are What we do Machine Learning Bio-i What we look for 20 20
Linear inseparability (in 3D) Who we are What we do Machine Learning Bio-i What we look for 21 21
Linear inseparability (in 3D) Who we are What we do Machine Learning Bio-i What we look for 21 21
Remember ... Goal: Characterise unknown toxic compound by means of genetic expression fingerprint(s) 22 22
Remember ... Goal: Characterise unknown toxic compound by means of genetic expression fingerprint(s) 22 22
Gene expression Who we are What we do Machine Learning Bio-i What we look for 23
Gene expression Who we are What we do Machine Learning Bio-i What we look for 23
Gene expression Who we are What we do Machine Learning Bio-i What we look for 23
Gene expression Who we are What we do Machine Learning Bio-i What we look for 23
Gene expression Who we are What we do Machine Learning Bio-i What we look for 23
Microarray technology Who we are What we do Machine Learning Bio-i What we look for 24
Infering transcriptional regulatory networks Who we are What we do Machine Learning Bio-i What we look for microarray causal relations data between genes 25
A transcriptional regulatory network Who we are What we do Machine Learning Bio-i What we look for 26
A transcriptional regulatory network Who we are What we do Machine Learning Bio-i What we look for 27
Systems Biology Systems biology is an academic Who we are field that seeks to integrate different levels of information to What we do understand how biological systems Machine function. Learning Bio-i What we look for 28
Systems Biology Systems biology is an academic Who we are field that seeks to integrate different levels of information to What we do understand how biological systems Machine function. Learning Bio-i What we look for 28
Systems Biology Systems biology is an academic Who we are field that seeks to integrate different levels of information to What we do understand how biological systems Machine function. Learning Bio-i transcriptome What we look for 28
Systems Biology Systems biology is an academic Who we are field that seeks to integrate different levels of information to What we do understand how biological systems Machine function. Learning Bio-i transcriptome proteome What we look for 28
Recommend
More recommend