Dr Nikolay Nikolaev

Full information in Nikolay's homepage

Areas of supervision

Evolutionary computation, genetic algorithms & genetic programming, neural networks, biocomputation, machine learning, applications to time-series prediction, financial engineering & data mining.

Featured works

Book chapters and conference papers

Nikolaev,N., and Iba,H. (2002). Genetic Programming of Polynomial Models for Financial Forecasting. In: Shu-Heng Chen (Ed.), Genetic
Algorithms and Genetic Programming in Computational Finance, Chapter 5, Kluwer Academic Publ., Boston, MA, pp.103-123.

Nikolaev,N., de Menezes,L. and Iba, H. (2002). Overfitting Avoidance in Genetic Programming of Polynomials, In: Proc. 2002 Congress on
Evolutionary Computation, CEC2002, IEEE Press, Piscataway, NJ, pp.1209-1214.

Nikolaev,N. and Iba, H. (2001). Genetic Programming using Chebishev Polynomials, In: L.Spector, E.D.Goodman, A.Wu, W.B.Langdon,
H.-M.Voigt, M.Gen, S.Sen, M.Dorigo, S.Pezeshk, M.H.Garzon, and E.Burke (Eds.), Proc. of the Genetic and Evolutionary Computation
Conference, GECCO-2001, Morgan Kaufmann Publ., San Francisco, CA, pp.89-96.

Publications and research outputs

Conference or Workshop Item

• Standard feedforward neural networks with backprop cannot support cognitive superposition

Article

• A Regime-Switching Recurrent Neural Network Model Applied to Wind Time Series
• Time-dependent series variance learning with recurrent mixture density networks
• Heavy-tailed mixture GARCH volatility modeling and Value-at-Risk estimation
• Nonlinear maximum likelihood estimation of electricity spot prices using recurrent neural networks
• Time-Dependent Series Variance Estimation via Recurrent Neural Networks
• Efficient online recurrent connectionist learning with the ensemble Kalman filter
• Recursive Bayesian Recurrent Neural Networks for Time-Series Modeling
• Single-Stacking Conformity Approach to Reliable Classification
• k-Version-Space Multi-class Classification Based on k-Consistency Tests
• Sequential Bayesian Kernel Modelling with Non-Gaussian Noise
• Polynomial Harmonic GMDH Learning Networks for Time Series Modeling
• Learning polynomial feedforward neural networks by genetic programming and backpropagation
• Genetic Programming of Polynomial Harmonic Networks using the Discrete Fourier Transform
• Accelerated Genetic Programming of Polynomials
• Regularization Approach to Inductive Genetic Programming

Book Section

• Nonlinear filtering of asymmetric stochastic volatility models and Value-at-Risk estimation

Book

• Adaptive Learning of Polynomial Networks: Genetic Programming, Backpropagation and Bayesian Methods

Research Interests

Neural networks
statistical learning networks, basis-function networks, constructive learning of the topology and initial weights of
multilayer neural networks; financial engineering by basis-function neural networks; chaotic time-series prediction by
statistical networks.

Genetic Algorithms
Structured genetic algorithms with cooperative subpopulations flowing on fitness sublandscapes; Fourier expansions
of fitness landscapes over regular graphs, messy genetic algorithms for applied economic regression tasks.

Inductive Genetic Programming (iGP):
Evolutionary induction of multivariate high-order polynomials, genetic programming of statistical learning networks,
genetic programming of polynomial discriminant classifiers, regularization in iGP, finite-state automata induction by
iGP.
Data mining
A utomated discovery of polynomials from data with numerical and continuous features; sequential forward and
backward feature selection for construction of multi-layer neural networks.

Machine Learning
Decision tree classifiers, stochastic complexity (Minimum Description Length-MDL) measures for decision tree
learners, multivariate splitting methods for non-linear decision trees; linear and oblique decision trees,
distance-based decision trees.

Current research

My recent work is devoted to genetic programming of tree-structured polynomials, known as statistical learning networks of the
GMDH type. This includes design of stochastic complexity (Minimum Description Length-MDL) and statistical
fitness functions for efficient search navigation. These functions are elaborated using ideas from the
regularization theory aiming at evolution of parsimonious, accurate and predictive polynomials.