Pavel marmalyuk

Classification methods: decision trees, random forests,Bayesian approach, discriminative functions, LDA, logistic regression, multilayer perceptrone (backpropagation), kernel methods, neural networks on radial-basis functions, nearest neighbours (e.g. condensed), model ensembles (bagging, boosting, cascading).

Regression methods: regularised linear regression, polynomial regression, kernel-based linear regression, regression trees.

Dimensionality reduction: LDA, PCA, factor analysis and structural equations modelling, latent trait modelling, wavelet transforms.

Signal processing: Fourier transform, wavelet transforms, denoising, decomposition and forecasting.

Image processing: simple pattern-matching techniques, connected components labeling algorithm, familiarity with complex methods such as Hough transform and Viola-Jones method.

Cluster analysis of data with vector or tensor structure (signals, surfaces).

Feature extraction and selection: filtering and weighting approaches, PCA.

Recommender Systems: basic K-NN based on Pearson correlation; rating matrix factorisation using gradient descent.

Model selection and validation: K-fold cross-validation, bootstrap techniques, Monte-Carlo techniques, learning curves, manual error analysis.

Algorithm performance analysis: R code profiling, time and memory consumption simple analysis techniques. 

Solid experience of R programming using RStudio IDE: S3/S4-classes, generic functions, vectorisation techniques, base/lattice/ggplot2 graphical systems, data analysis (dplyr, data.table, reshape, lubridate, tm, mlr, caret, signal, ts, MASS, etc), reports (knitr), interactive dashboards (Shiny), parallel computing (foreach) .

Software developed:

- software for financial risk assessment and control (energy sector) based on the Monte-Carlo method utilizing Geometric Brownian Motion model;

- tools for analysis of large amounts of sensor data (~20 Gb) for chemical production company: time series preprocessing, feature engineering and selection, exploratory analysis, anomaly detection and machine learning (Extremely Randomized Trees techinque for prediction of product quality);

- system for analysis of videooculography data (smoothing, segmentation, clustering, kernel density estimation,  multivariate statistical analysis, complex visualisations);

- utilities for revealing of temporal patterns in psychological testing process by analysis of logged assessment history;

- implementation of Markov chains' parameters optimisation techniques (gradient descent, genetic algorithm, combinatorial methods);

- utilities for calculating images similarity metrics based on 2D-wavelet  transform and linear regression analysis allowing revealing of details in images that influence subjects choice in face discrimination task;

- utilities for estimation of time series of observed state probabilities needed for Markov chains parameters identification;

- software for optimisation of parameters of discriminant functions applied for binary classification tasks;

- software for SMS spam detection using naive Bayes classifier utilizing keywords' probabilities estimates;

- software for polynomial regression fitting.

Solid experience of LabVIEW programming for developing data analysis products (mostly signal and image processing and Markov chains optimisation).

Software developed:

- mobile application for blind users allowing detection of obstacles utilizing acoustic sonar principle;

- FPGA-software controlling autonomous robot;

- system for biomechanical data acquisition, processing (spectral analysis, wavelet transforms, statistics estimation) and reporting;

- pattern-matching software detecting light-reflective markers in video frames;

- pattern-matching software detecting pathological segments in stabilographical signals utilizing relaxation neural network and wavelet coefficients;

- software for confirmatory factor models validation using Kohonen's self-organising maps and Monte-Carlo approach;

- classification software for estimation of probabilities of aircraft pilots proficiency levels using probabilisitc neural network (similar to kernel-weighted K-nearest neighbour approach);

- implementation of Markov chains' parameters optimisation techniques (gradient descent, genetic algorithm, own combinatorial methods).

More than 70 articles and abstracts published, including 7 articles in internaltional journals (see http://elibrary.ru/author_items.asp?authorid=616227).

Presentation of various projects results at more than 40 conferences (including 5 international European conferences), 15 domestic exhibitions.

Moderate experience (3 years) of giving lectures and practical lessons in the field of Computer Science. More than 10 successfull diploma and 20 course projects.

Experience of using IBM SPSS and StatSoft Statistica for samples comparison, distributions fitting, correlation, regression, factor, LTM and discriminant analysis for educational needs: psycho-educational questionaries standartisation, validation and data mining using small datasets.

DFD, ER diagrams creation for small software development projects in the area of experimental and cognitive psychology.

Moderate experience of MS Excel for data analytics (probability and statistics, visualisation, Data Analysis and Solver add-ins).

Writing PL-SQL queries and stored procedures for Oracle 9i database. Experience using all relational operations, including JOINs and nested queries.

Moderate experience in using Shiny web application framework: developing local application for analysis of videooculography data (see http://github.com/PMarmalyuk/EyeTrackingProject).

Theoretical background of the Map Reduce computational model, HDFS, Hadoop stack representative developer tools.

Elementary experience of C++ programming during Master's degree university program.

Little Matlab/Octave experience: writing basic functions for machine learning systems developed during the Andrew Ng's Machine Learning course at Coursera.

Elementary experience of C# programming: small modification of GazeTracker system to allow to use Sony PS3Eye Camera with the system  (see http://sourceforge.net/projects/gazetrackinglib/files/gazetrackinglib/).