Computer Science
Loading...
ISSN 1508-2806
e-ISSN: 2300-7036
Issue Date
2020
Volume
Vol. 21
Number
No. 4
Description
Reviewed by: Maciej Wozniak, Janusz Szczepanski, Eirik Valseth, Joao Saraiva, Astik Biswas, Piotr Szwed, Vishal Pasricha, Anna Paszynska, Csaba Szabo, Omer Faruk Alis, Maciej Paszynski
Journal Volume
Computer Science
Vol. 21 (2020)
Projects
Pages
Articles
Hybrid CNN-Ligru acoustic modeling using sincnet raw waveform for hindi ASR
(Wydawnictwa AGH, 2020) Kumar, Ankit; Aggarwal, Rajesh Kumar
Deep neural networks (DNN) currently play a most vital role in automatic speech recognition (ASR). The convolution neural network (CNN) and recurrent neural network (RNN) are advanced versions of DNN. They are right to deal with the spatial and temporal properties of a speech signal, and both properties have a higher impact on accuracy. With its raw speech signal, CNN shows its superiority over precomputed acoustic features. Recently, a novel first convolution layer named SincNet was proposed to increase interpretability and system performance. In this work, we propose to combine SincNet-CNN with a light-gated recurrent unit (LiGRU) to help reduce the computational load and increase interpretability with a high accuracy. Different configurations of the hybrid model are extensively examined to achieve this goal. All of the experiments were conducted using the Kaldi and Pytorch-Kaldi toolkit with the Hindi speech dataset. The proposed model reports an 8.0% word error rate (WER).
Comparison of multi-frontal and alternating direction parallel hybrid memory iGRM direct solver for non-stationary simulations
(Wydawnictwa AGH, 2020) Woźniak, Maciej; Bukowska, Anna
Three-dimensional isogeometric analysis (IGA-FEM) is a modern method for simulation. The idea is to utilize B-splines or NURBS basis functions for both computational domain descriptions and engineering computations. Refined isogeometric analysis (rIGA) employs a mixture of patches of elements with B-spline basis functions and $C^0$ separators between them. This enables a reduction in the computational cost of direct solvers. Both IGA and rIGA come with challenging sparse matrix structures that are expensive to generate. In this paper, we show a hybrid parallelization method using hybrid-memory parallel machines. The two-level parallelization includes the partitioning of the computational mesh into sub-domains on the first level (MPI) and loop parallelization on the second level (OpenMP). We show that the hybrid parallelization of the integration reduces the contribution of this phase significantly. We compare the multi-frontal solver and alternating direction solver, including the integration and the factorization phases.
Tunnel parsing with counted repetitions
(Wydawnictwa AGH, 2020) Handzhiyski, Nikolay; Somova, Elena
This article describes a new and efficient algorithm for parsing (called tunnel parsing) that parses from left to right on the basis of context-free grammar without left recursion nor rules that recognize empty words. The algorithm is mostly applicable for domain-specific languages. In the article, particular attention is paid to the parsing of grammar element repetitions. As a result of the parsing, a statically typed concrete syntax tree is built from top to bottom, that accurately reflects the grammar. The parsing is not done through a recursion, but through an iteration. The tunnel parsing algorithm uses the grammars directly without a prior refactoring and is with a linear time complexity for deterministic context-free grammars.
Forecasting currency exchange rate time series with fireworks algorithm-based higher order neural network, with special attention to training data enrichment
(Wydawnictwa AGH, 2020) Sahu, Kishore Kumar; Nayak, Sarat Chandra; Behera, Himansu Sekhar
Exchange rates are highly fluctuating by nature, thus, they are difficult to forecast. Artificial neural networks (ANNs) have proven to be better than statistical methods. Inadequate training data may lead the model to reach sub-optimal solutions, resulting in poor accuracy (as ANN-based forecasts are data-driven). To enhance forecasting accuracy, we suggests a method of enriching training datasets through exploring and incorporating virtual data points (VDPs) by an evolutionary method called the fireworks algorithm-trained functional link artificial neural network (FWA-FLN). The model maintains a correlation between current and past data, especially at the oscillation point on the time series. The exploration of a VDP and forecast of the succeeding term go consecutively by FWA-FLN. Real exchange rate time series are used to train and validate the proposed model. The efficiency of the proposed technique is related to other similarly trained models and produces far better prediction accuracy.
Causal reversibility in individual token interpretation of Petri Nets
(Wydawnictwa AGH, 2020) Benamira, Adel
Causal reversibility in concurrent systems means that events that the origin of other events can only be undone after undoing its consequences. In opposition to backtracking, events that are independent of each other can be reversed in an arbitrary order, in other words, we have flexible reversibility with respect to a causality relationship. An implementation of individual token interpretation of Petri Nets (IPNs) has been proposed by Rob Van Glabbeek et al., the present paper investigates a study of causal reversibility within IPNs. Given N as an IPN, by adding an intuitive firing rule to undo transitions according to the causality relationship, the coherence of N is assured, i.e., the set of all reachable states of N in the reversible version and that of the original one are identical. Furthermore, reversibility in N is flexible, and their initial state can be accessible in reverse from any state. In this paper, an approach for controlling causal-reversibility within IPNs is proposed.