计算机技术国际实验室
简介
International Scientific Laboratory " Computer technology " is based on departments " Computer technology ", " Programming Technology ", " Software Engineering and program verification " of ITMO University, on " Algorithms assembly of genomic sequences " laboratory , created on the basis of decisions of the Academic Council of ITMO University (27.12 . 2011), and research center "Programming technologies and artificial intelligence ", organized in the framework of the program of ITMO University (2009-2018) , and scientific and educational center " development of genome assembly methods , transcriptome assembly and proteome dynamic analysis" established under the Federal Target Program " scientific and scientific-pedagogical personnel of innovative Russia " for 2009-2013 . Action 1.1 "Support of research pursued by the research and education centers " in a scientific direction " Life Sciences ( Living Systems ) " in " Genomic , proteomic and post-genome technologies".
Laboratory conducts research in four areas: coding theory, bioinformatics, machine Learning, software engineering.
研究方向
Bioinformatics
Computer Technologies lab work in three areas of bioinformatics:
- sequencing data analysis;
- expression data analysis;
- integrating different types of biological data.
In sequencing data analysis we work on developing mathematical models and algorithms for genome assembly. It is a very important task because assembled genomes are used in a variety of bioinformatical pipelines (e.g. variant calling). We also study the problem of genome assembly from theoretical point of view. For instance, we have prooved NP-hardness of the genome assembly problem.
Expression data analysis includes designing methods of extracting biologicaly relevant information from gene expression profiles. It could be, for example, finding a set of genes characteristic for a particular type of cells. Another problem we are working on is gene expression deconvolution: determining what type of cells in what
quantities are present in samples.
Advance in biological profiling technologies also allows to use not only one technology (for exampla, RNA-sequencing) but mutltiple of them. This poses a problem of integrating these multiple types of data to view on the biological problem simulthaneously from multiple perspectives. Currently, in a collaboration with Washington University in St. Louis we work on integrating transcriptional and metabolic profiling data in a context of biochemical reactions.
Machine Learning
Machine learning methods are developed, including methods for the analysis of big data. In particular, methods for classification of users of online resources are proposed.
Methods for solving machine learning problems are being developed using SAT-solvers and CSP-solvers, as well as swarm intelligence techniques. Theoretical study of evolutionary algorithms is being performed, including time analysis of evolutionary algorithms adjusted with reinforcement learning.
ADAPTIVE SELECTION OF EXTRA OBJECTIVES USING REINFORCEMENT LEARNING
Single-objective optimization can be enhanced by adding auxiliary objectives, but how should we choose the most efficient ones, and when should we use the particular objective? A method designed to solve these issues was proposed in our laboratory [1-3]. The method is called EA+RL, which stands for an evolutionary algorithm (EA) controlled with reinforcement learning (RL).
There are several techniques that involve using some additional objectives in order to enhance performance of EAs. In multiobjectivization technique [6] all the objectives are optimized simultaneously by some multiobjective algorithms (MOEAs). In this technique the objectives should be specially developed in order to increase the optimization performance. It was shown that adding an inefficient objective leads MOEAs to fail on the considered model problems [3].
Helper-objective approach also involves using MOEAs, but it requires a strategy of choosing the auxiliary objective to be optimized at the current population [5]. The strategy can be either random, or ad-hoc [7]. The random one is general, but it does not take advantage of problem characteristics. At the same time, ad-hoc strategies can be efficient, but they lack generality.
The EA+RL method incorporates auxiliary objectives into a single-objective EA. It requires less computational effort than MOEA-based methods, which makes it more applicable to resource-consuming problems. The selection strategy used in EA+RL is problem independent and it allows to learn some features of the problem as well, thus the method seems to increase both efficiency and generality of the helper-objective approach.
There are several works that investigate the use of RL for adjustment of EAs. In some of them, tuning of numerical parameters, such as mutation probability and population size, is considered [4, 8], in other papers, evolutionary operators selection [9, 10] is investigated. Using RL as a strategy of choosing auxiliary objectives in EAs was proposed in the EA+RL method for the first time.
The EA+RL method of selecting fitness function of an evolutionary algorithm using reinforcement learning, t is the function to be optimized, i is the number of the current generation.
Programming Technologies
All software developers dream of the times when computers will write programs for them, or, if it is not possible, that computers will at least help them in this hard task. A large variety of software projects, from assemblers, compilers and interpreters to UML-based executable code generators are aimed at reaching this goal.
Search-based software engineering (SBSE) is a collection of techniques for solving problems that arise in the process of software development by translating them to optimization problems and, consequently, solving these problems. SBSE uses optimization methods such as genetic and evolutionary algorithms, ant colony optimization and other stochastic search algorithms.
The goal in the “Programming technologies” research direction of the Computer Technologies Laboratory is to create SBSE methods that would allow to automate the processes of design, development and testing of software.
联系方式
VLADIMIR GLEBOVICH PARFENOV
教授、科学博士