Игнатов Дмитрий Игоревич

Identifying dyslexia in school pupils from eye movement and demographic data using artificial intelligence

2023 · ARTICLE · en

This paper represents our research results in the pursuit of the following objectives: (i) to introduce a novel multi-sources data set to tackle the shortcomings of the previous data sets, (ii) to propose a robust artificial intelligence-based solution to identify dyslexia in pri- mary school pupils, (iii) to investigate our psycholinguistic knowledge by studying the impor- tance of the features in identifying dyslexia by our best AI model. In order to achieve the first objective, we collected and annotated a new set of eye-movement-during-reading data. Fur- thermore, we collected demographic data, including the measure of non-verbal intelligence, to form our three data sources. Our data set is the largest eye-movement data set globally. Unlike the previously introduced binary-class data sets, it contains (A) three class labels and (B) reading speed. Concerning the second objective, we formulated the task of dyslexia pre- diction as regression and classification problems and scrutinized the performance of 12 classifications and eight regressions approaches. We exploited the Bayesian optimization method to fine-tune the hyperparameters of the models: and reported the average and the standard deviation of our evaluation metrics in a stratified ten-fold cross-validation. Our studies showed that multi-layer perceptron, random forest, gradient boosting, and k-nearest neighbor form the group having the most acceptable results. Moreover, we showed that although separately using each data source did not lead to accurate results, their combina- tion led to a reliable solution. We also determined the importance of the features of our best classifier: our findings showed that the IQ, gender, and age are the top three important fea- tures; we also showed that fixation along the y-axis is more important than other fixation data.

Towards interaction-based user embeddings in sequential recommender models

2022 · CHAPTER · en

All transductive recommender systems are unable to make predictions for users who were not included in the training sample due to the process of learning user-specific embeddings. In this paper, we propose a new method for replacing identity-based user embeddings in existing sequential models with interaction-based user vectors trained purely on interaction sequences. Such vectors are composed of user interactions using GRU layers with adjusted dropout and maximum item sequence length. This approach is substantially more efficient and does not require retraining when new users appear. Extensive experiments on three open-source datasets demonstrate noticeable improvement in quality metrics for the most of selected state-of-the-art sequential recommender models.

DaNetQA: a yes/no Question Answering Dataset for the Russian Language

2021 · CHAPTER · en

Efficient Language Modeling with Automatic Relevance Determination in Recurrent Neural Networks

2019 · CHAPTER · en

Reduction of the number of parameters is one of the most important goals in Deep Learning. In this article we propose an adaptation of Doubly Stochastic Variational Inference for Automatic Relevance Determination (DSVI-ARD) for neural networks compression. We find this method to be especially useful in language modeling tasks, where large number of parameters in the input and output layers is often excessive. We also show that DSVI-ARD can be applied together with encoder-decoder weight tying allowing to achieve even better sparsity and performance. Our experiments demonstrate that more than 90% of the weights in both encoder and decoder layers can be removed with a minimal quality loss.

Predicting Winning Team and Probabilistic Ratings in Dota 2 and Counter-Strike: Global Offensive Video Games

2018 · CHAPTER · en

In this paper, we present novel winning team predicting models and compare the accuracy of the obtained prediction with TrueSkill model of ranking individual players impact based on their impact in team victory for the two most popular online games: Dota 2 and Counter-Strike: Global Offensive.

Realistic post-processing of rendered 3D scenes

2018 · CHAPTER · en

In this talk, we show a realistic post-processing rendering based on generative adversarial network CycleWGAN. We propose to use CycleGAN architecture and Wasserstein loss function with additional identity component in order to transfer graphics from Grand Theft Auto V to the older version of GTA video-game, Grand Theft Auto: San Andreas. We aim to present the application of modern art style transfer and unpaired image-to-image translations methods for graphics improvement using deep neural networks with adversarial loss.

Online recommender system for radio station hosting based on information fusion and adaptive tag-aware profiling

2016 · ARTICLE · en

We present a new recommender system developed for the Russian interactive radio network FMhost. To the best of our knowledge, it is the first model and associated case study for recommending radio stations hosted by real DJs rather than automatically built streamed playlists. To address such problems as cold start, gray sheep, boosting of rankings, preference and repertoire dynamics, and absence of explicit feedback, the underlying model combines a collaborative user-based approach with personalized information from tags of listened tracks in order to match user and radio station profiles. This is made possible with adaptive tag-aware profiling that follows an online learning strategy based on user history. We compare the proposed algorithms with singular value decomposition (SVD) in terms of precision, recall, and normalized discounted cumulative gain (NDCG) measures; experiments show that in our case the fusion-based approach demonstrates the best results. In addition, we give a theoretical analysis of some useful properties of fusion-based linear combination methods in terms of graded ordered sets.

A Lattice-based Consensus Clustering Algorithm

2016 · CHAPTER · en

We propose a new algorithm for consensus clustering, FCA-Consensus, based on Formal Concept Analysis. As the input, the algorithm takes T partitions of a certain set of objects obtained by k-means algorithm after T runs from different initialisations. The resulting consensus partition is extracted from an antichain of the concept lattice built on a formal context objects×classes, where the classes are the set of all cluster labels from each initial k-means partition. We compare the results of the proposed algorithm in terms of ARI measure with the state-of-the-art algorithms on synthetic datasets. Under certain conditions, the best ARI values are demonstrated by FCA-Consensus.

Bayesian Learning of Consumer Preferences for Residential Demand Response

2016 · ARTICLE · en

Статья посвящена моделированию поведения пользователя хлебопечки как устройства умного дома. На основании реальных цен на электричество и симуляции поведения пользователя требуется предложить математическую модель и алгоритм машинного обучения для оптимального по соотношению цен и полезности для пользователя автоматического запуска устройства. Мы предлагаем Баейсовский алгоритм машинного обучения для обучения предпочтениям пользователя с учетом предудыщих запусков. В работе демонстрируются результаты экспериментов, в частности, превосходство поедлагаемого алгоритма над популярной реализацией градиентого бустинга, XGBoost. Подход может быть перенесен и на другие бытовые приборы, такие как системы обогрева и кондиционирования.

Preface

2016 · CHAPTER · en