arXiv subjects

Pavel Prikhodko

Publications and source records attributed to Pavel Prikhodko.

3 recordsLinked to original sources

arXiv2018

Emotion Recognition From Speech With Recurrent Neural Networks

In this paper the task of emotion recognition from speech is considered. Proposed approach uses deep recurrent neural network trained on a sequence of acoustic features calculated over small speech intervals. At the same time special probabilistic-nature CTC loss function allows to consider long utterances containing both emotional and neutral parts. The effectiveness of such an approach is shown in two ways. Firstly, the comparison with recent advances in this field is carried out. Secondly, human performance on the same task is measured. Both criteria show the high quality of the proposed method.

cs.CL↗

arXiv2018

Calibration of Sobol indices estimates in case of noisy output

This paper presents a simple noise correction method for Sobol' indices estimation. Sobol' indices, especially total Sobol' indices are quite sensitive to the noise in the output and tend to be severly biased (overestimated) if no noise correction is done, which may make their computation meaningless in case of even quite moderate noise levels. Proposed method allows to get approximately unbiased noise free estimation of Sobol' indices at the cost of variance of estimate increase if noise can be represented as a combination of additive and multiplicative stationary noise. %Also we show that it is impossible to do precise noise correction for more complex noise settings. Proposed method is more straightforward than schemes found in the literature and does not introduce any assumptions on the function and noise distribution (except that it assumes noise to be stationary and be a combination of additive and multiplicative). One of the appealing features is that there is actual analytical noise correction expression derived.

stat.CO↗

arXiv2016

GTApprox: surrogate modeling for industrial design

We describe GTApprox - a new tool for medium-scale surrogate modeling in industrial design. Compared to existing software, GTApprox brings several innovations: a few novel approximation algorithms, several advanced methods of automated model selection, novel options in the form of hints. We demonstrate the efficiency of GTApprox on a large collection of test problems. In addition, we describe several applications of GTApprox to real engineering problems.

cs.MS↗