Pollack Periodica
Volume/Issue:
Volume 19: Issue 3
Exploring noise-induced techniques to strengthen deep learning model resilience
Authors:
Aadil Gani Ganie Department of Applied Informatics, Faculty of Mechanical Engineering and Informatics, Institute of Information Sciences, University of Miskolc, Miskolc-Egyetemváros, Hungary

Search for other papers by Aadil Gani Ganie in
Current site
Google Scholar
PubMed Close
https://orcid.org/0000-0002-6607-6994
and
Samad Dadvadipour Department of Applied Informatics, Faculty of Mechanical Engineering and Informatics, Institute of Information Sciences, University of Miskolc, Miskolc-Egyetemváros, Hungary

Search for other papers by Samad Dadvadipour in
Current site
Google Scholar
PubMed Close
Pages:
8–13
Online Publication Date:
06 May 2024
Publication Date:
16 Oct 2024
Article Category:
Research Article
DOI:
https://doi.org/10.1556/606.2024.01068
Keywords:
deep learning; noise-induced regularization; model generalization; long short-term memory; bidirectional long short-term memory
Open access

Abstract

In artificial intelligence, combating overfitting and enhancing model generalization is crucial. This research explores innovative noise-induced regularization techniques, focusing on natural language processing tasks. Inspired by gradient noise and Dropout, this study investigates the interplay between controlled noise, model complexity, and overfitting prevention. Utilizing long short-term memory and bidirectional long short term memory architectures, this study examines the impact of noise-induced regularization on robustness to noisy input data. Through extensive experimentation, this study shows that introducing controlled noise improves model generalization, especially in language understanding. This contributes to the theoretical understanding of noise-induced regularization, advancing reliable and adaptable artificial intelligence systems for natural language processing.

Abstract

In artificial intelligence, combating overfitting and enhancing model generalization is crucial. This research explores innovative noise-induced regularization techniques, focusing on natural language processing tasks. Inspired by gradient noise and Dropout, this study investigates the interplay between controlled noise, model complexity, and overfitting prevention. Utilizing long short-term memory and bidirectional long short term memory architectures, this study examines the impact of noise-induced regularization on robustness to noisy input data. Through extensive experimentation, this study shows that introducing controlled noise improves model generalization, especially in language understanding. This contributes to the theoretical understanding of noise-induced regularization, advancing reliable and adaptable artificial intelligence systems for natural language processing.

1 Introduction

The rapid progression of deep learning techniques has transformed numerous domains, including natural language processing and computer vision, enabling the creation of intricate and precise models [1]. However, the escalating complexity of these models raises concerns about their inclination to overfit the training data, impeding their adaptability to unseen data instances [2]. Regularization techniques are pivotal in mitigating this challenge, encouraging model generalization by discouraging excessively intricate solutions [3]. While traditional methods like L1 and L2 regularization have been extensively explored and applied [4], recent years have witnessed a paradigm shift towards considering noise as a regularization tool. The strategic introduction of controlled noise during training has displayed promising outcomes in bolstering model resilience and refining generalization performance [5].

This study delves into the innovative realm of noise-induced regularization within deep learning models, concentrating on Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (BiLSTM) architectures [6, 7]. By purposefully integrating Gaussian noise into the training procedure, the research delves into noise's impact on the models' learning dynamics and their capacity to generalize across diverse and noisy datasets [8]. The inquiry extends to the mathematical foundations of noise-induced regularization, illuminating the complex interplay among noise, model intricacy, and generalization abilities [9]. Within this context, this paper surveys existing literature on conventional regularization methods and noise-induced regularization, providing a holistic view of the evolution of regularization techniques in deep learning.

Leveraging recent investigations that have explored noise as a regularization mechanism [10], this study presents a methodical analysis of its influence on training dynamics and model efficacy. By amalgamating theoretical insights with empirical discoveries, this research endeavors to establish a profound comprehension of noise-induced regularization and its potential applications in augmenting the resilience of deep learning models.

2 Literature review

2.1 Regularization techniques in deep learning

Essential to deep learning models, regularization techniques significantly enhance their ability to generalize. Conventional methods like L1 and L2 regularization curb overfitting by penalizing large parameter values [4]. Another popular approach, dropout, introduced by Srivastava et al. [10], involves randomly dropping units during training to prevent co-adaptation of feature detectors [4]. Recent innovations have introduced new regularization methods, including batch normalization [11] and weight regularization [12], proving effective in optimizing model performance.

2.2 Noise-induced regularization in deep learning

In recent years, noise-induced regularization has emerged as a promising strategy to bolster model generalization. Neelakantan et al. [5] pioneered the concept of adding gradient noise, enhancing the learning dynamics of deep networks [13]. Vincent et al. [8] proposed denoising autoencoders, which learn robust features by reconstructing clean inputs from noisy data. Dhifallah and Lu [14] harnessed noise injection as a regularization technique, enhancing the robustness of convolutional neural networks in image recognition tasks [5]. These studies underscore the potential of noise-induced regularization in fortifying deep learning models against noisy input data.

2.3 Application of regularization techniques in natural language processing

In the realm of Natural Language Processing (NLP), regularization techniques have significantly enhanced task performance. Vincent et al. [8] investigated the impact of regularization methods on machine translation tasks, demonstrating L2 regularization's effectiveness in enhancing translation accuracy. Furthermore, recent studies have explored noise-induced regularization specifically in NLP tasks. K. Zhang et al. [15] introduced noise to word embeddings, enhancing sentiment analysis accuracy. A. Pretorius et al.. [16] applied noise-induced regularization to recurrent neural networks, improving text generation model performance.

2.4 Noise-induced regularization in LSTM and BLSTM models

The application of noise-induced regularization in LSTM and BiLSTM models has gained traction. M. Qiao et al. [17] introduced noise to LSTM networks' input sequences, resulting in improved performance in sequence prediction tasks. Similarly, A. A. Abdelhamid et al. [18] incorporated noise into the training process of BiLSTM models, enhancing their ability to capture complex dependencies in sequential data. These studies highlight noise-induced regularization's potential in LSTM and BiLSTM [19] architectures for diverse sequential tasks. In summary, while traditional methods provide a foundation, recent strides in noise-induced regularization offer promising avenues to enhance model robustness, especially in the context of LSTM and BiLSTM architectures for sequential tasks in NLP and beyond.

3 Results and discussions

3.1 Methodology

Data preparation: A dataset comprising text data with “suicide” and “non-suicide” labels was preprocessed using NLP techniques, including tokenization and lemmatization.

Model architectures: LSTM and BiLSTM architectures were chosen for their effectiveness in sequential data tasks.

Noise-induced regularization: Controlled Gaussian noise was injected into the input data to induce regularization, preventing overfitting and enhancing the models' robustness as it is shown in Fig. 1.

Fig. 1.
Fig. 1.

Proposed workflow within the use of noise induced regularization

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Training and optimization: The models were trained using Adam optimizer with backpropagation, optimizing a composite objective function that integrated noise-induced regularization.

Evaluation: Model performance was evaluated based on accuracy scores under varying noise levels, demonstrating the effectiveness of noise-induced regularization in improving generalization capabilities.

3.2 Mathematical optimization

This section, delves into the mathematical formulations, optimization strategies, and experimental results of noise-induced regularization experiments on LSTM and BiLSTM models. These experiments were conducted on a dataset comprising text data with two distinct classes, utilizing techniques from the field of NLP. The LSTM model processes input sequences xt and maintains a hidden state ht and a cell ct. The equations for the LSTM are,
it=σ(Wiixt+bii+Whiht−1+bhi),
ft=σ(Wifxt+bif+Whfht−1+bhf),
gt=tanh(Wigxt+big+Whght−1+bhg),
ot=σ(Wioxt+bio+Whoht−1+bho),
ct=ft.ct−1+it.gt,
ht=ot·tanhct,
where xt is the input at time step t; it,ft,gt,ot are the input, forget, cell and output gates, respectively; σ represents the sigmoid activation function, b is the bias and W weights.
The objective function for the LSTM model, incorporating noise-induced regularization, is defined as:
minθllstm(θ)+λ·NoiseRegularizationθ,σ.
Here, θ represents the model parameters, llstm(θ) denotes the standard LSTM loss function applied to the NLP task, λ controls the regularization strength, and NoiseRegularization(θ,σ) quantifies the noise-induced regularization term, where σ signifies the noise level. Bidirectional LSTM processes the input sequence in both forward and backward directions. The equations for BiLSTM and similar to LSTM, but it processes the input in reverse order as well,
h1t=LSTMforward(h1t−1,xt),
h2t=LSTMbackward(h2t+1,xt),
ht=[h1t+h2t],
where h1t and h2t are the hidden states of forward and backward LSTM's respectively.
The objective function for both LSTM and BiLSTM models can be defined as a combination of a loss term and a regularization term to handle noise. Considering Mean Squared Error (MSE) as the loss function, the objective function (l) can be defined as follows:
l=MSEyˆ,ttrue+λ·RegularizationTerm,
where yˆ represents the predicted output; ytrue represents the true output; λ controls the strength of the regularization term; RegularizationTerm can be a measure of the noise or complexity in the model, e.g., the L2 norm of the parameters or the variance of the noise added.
To optimize the objective function with the added noise term, the noise matrix n is considered as part of the input to the model. The objective function with noise was the regularization term captures, the noise introduced into the model. To optimize this objective function, the partial derivative with respect to the model parameters θ are taking and considering both the prediction error and the noise regularization term. Taking the partial derivative of l with respect to θ yields:
∂l∂θ=∂MSEyˆ,ytrue∂θ+λ·∂RegularizationTerm∂θ.
Now, let's consider RegularizationTerm = ‖n‖2/2 where ‖n‖2 represents the squared L2 norm of the noise matirx∙n. The regularization term encourages the model to learn robust features that are not overly dependent on noisy input. The partial derivation of the regularization term with respect to θ is λ·n, assuming n is treated as a constant during differentiation. Therefore, the overall partial derivative of the objective function with respect to θ becomes
∂l∂θ=∂MSEyˆ,ytrue∂θ+λ·n.

This partial derivative has been used in gradient-based optimization algorithm Adam in this case to update the model parameters θ during training. The regularization term λ∙n penalizes the model for being sensitive to the noise introduction, leading to more generalized and robust learning.

The dataset utilized in these experiments consists of text data with two distinct classes: “suicide” and “non-suicide.” This dataset was preprocessed using various NLP techniques, including tokenization, stopword removal, and lemmatization, to prepare the text data for model training.

Figures 2–9 illustrates the accuracy scores for both LSTM and BiLSTM models under varying noise levels 0, 0.2, 0.5, and 0.7, respectively. In this study Gaussion noise has been added, which is a kind of signal noise that has a probability density function equal to that of the normal distribution. The figures reveal that, with the introduction of controlled noise, the training accuracy initially decreases due to increased complexity. However, the testing accuracy remains stable, showcasing the noise-induced regularization's efficacy in preventing overfitting and enhancing the models' generalization capability for NLP tasks.

Fig. 2.
Fig. 2.

LSTM model loss and accuracy at noise = 0.0

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 3.
Fig. 3.

LSTM model accuracy and loss at noise = 0.2

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 4.
Fig. 4.

LSTM model accuracy and loss at noise = 0.5

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 5.
Fig. 5.

LSTM model accuracy and loss at noise = 0.7

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 6.
Fig. 6.

BiLSTM model accuracy and loss at noise = 0.0

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 7.
Fig. 7.

BiLSTM model accuracy and loss at noise = 0.2

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 8.
Fig. 8.

BiLSTM model accuracy and loss at noise = 0.5

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 9.
Fig. 9.

BiLSTM model accuracy and loss at noise = 0.7

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

In the polynomial regression experiments, aimed at demonstrating the impact of different regularization techniques on overfitting. Figures 10–12 presents the results. These figures show cases the overfitting phenomenon with a 9th-degree polynomial, the effectiveness of dropout-like regularization with a 3rd-degree polynomial, and the noise-induced regularization's ability to mitigate overfitting with a 7th-degree polynomial.

Fig. 10.
Fig. 10.

Overfitting demonstration

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 11.
Fig. 11.

Demonstration of dropout-like regularization

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

Fig. 12.
Fig. 12.

Demonstration of the effect of noise

Citation: Pollack Periodica 19, 3; 10.1556/606.2024.01068

These results underscore the vital role of noise-induced regularization in enhancing model generalization for text data in the domain of NLP. By incorporating noise and leveraging NLP techniques, this approach ensures the models' robustness to noisy and complex language patterns, contributing to the advancement of natural language understanding in artificial intelligent systems.

4 Conclusion

In conclusion, this study has delved into the innovative realm of noise-induced regularization techniques within deep learning models, with a specific focus on LSTM and BiLSTM architectures. By systematically investigating the impact of controlled Gaussian noise on model learning dynamics and generalization capabilities, this research has contributed significantly to the understanding of noise-induced regularization and its applications in enhancing the resilience of deep learning models, particularly in NLP tasks.

4.1 Summary of research results

Throughout this experimentation, it is observed a consistent trend where the introduction of controlled noise during training led to improvements in model generalization, particularly evident in language understanding tasks. Findings demonstrate that despite the initial decrease in training accuracy due to increased complexity induced by noise, the testing accuracy remains stable, underscoring the efficacy of noise-induced regularization in preventing overfitting and enhancing model robustness to noisy input data.

4.2 Main added value of the research

The primary contribution of this research lies in its comprehensive exploration of noise-induced regularization techniques, extending the traditional understanding of regularization methods in deep learning. By elucidating the intricate interplay between noise, model complexity, and generalization abilities, this study advances the theoretical foundations of noise-induced regularization, paving the way for the development of more reliable and adaptable artificial intelligence systems, particularly in the domain of NLP. Furthermore, our investigation into the application of noise-induced regularization in LSTM and BiLSTM architectures enriches the existing literature by providing insights into their effectiveness for diverse sequential tasks in NLP and beyond.

4.3 Future directions

Moving forward, future research endeavors can build upon the findings of this study by exploring optimal integration strategies for traditional and noise-induced regularization methods, with a particular focus on their transferability to a broader range of deep learning tasks. Additionally, investigating noise-induced regularization in emerging architectures and real-world applications holds promise for further advancements in building robust and resilient deep learning models. In essence, this research not only contributes to the theoretical understanding of noise-induced regularization but also offers practical insights that can potentially drive advancements in the development of more adaptive and reliable deep learning models for various applications, particularly in the domain of natural language processing.

Acknowledgments

The Authors would like to acknowledge the support and guidance provided by the mentors and colleagues throughout this research project.

References

  • [1]↑

    Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, pp. 436–444, 2015.

  • [2]↑

    C. M. Bishop, Pattern Recognition and Machine Learning. Springer, 2006.

  • [3]↑

    S. Ruder, “An overview of gradient descent optimization algorithms,” arXiv:1609.04747, 2016.

  • [4]↑

    I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. MIT Press, 2016.

  • [5]↑

    A. Neelakantan, L. Vilnis, Q. V. Le, I. Sutskever, L. Kaiser, K. Kurach, and J. Martens, “Adding gradient noise improves learning for very deep networks,” arXiv:1511.06807, 2015.

    • Search Google Scholar
    • Export Citation
  • [6]↑

    A. Graves and J. Schmidhuber, “Framewise phoneme classification with bidirectional LSTM and other neural network architectures,” Neural Netw., vol. 18, nos 5–6, pp. 602–610, 2005.

    • Search Google Scholar
    • Export Citation
  • [7]↑

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Comput., vol. 9, no. 8, pp. 1735–1780, 1997.

  • [8]↑

    P. Vincent, H. Larochelle, Y. Bengio, and P. A. Manzagol, “Extracting and composing robust features with denoising autoencoders,” in Proceedings of the 25th International Conference on Machine Learning, Helsinki, Finland, July 5–9, 2008, pp. 1096–1103.

    • Search Google Scholar
    • Export Citation
  • [9]↑

    Q. Zheng, M. Yang, J. Yang, Q. Zhang, and X. Zhang, “Improvement of generalization ability of deep CNN via implicit regularization in two-stage training process,” IEEE Access, vol. 6, pp. 15844–15869, 2018.

    • Search Google Scholar
    • Export Citation
  • [10]↑

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: a simple way to prevent neural networks from overfitting,” J. Machine Learn. Res., vol. 15, no. 1, pp. 1929–1958, 2014.

    • Search Google Scholar
    • Export Citation
  • [11]↑

    S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proceedings of the 32nd International Conference on International Conference on Machine Learning, Lille, France, July 6–11, 2015, pp. 448–456.

    • Search Google Scholar
    • Export Citation
  • [12]↑

    T. van Laarhoven, “L2 regularization versus batch and weight normalization,” arXiv:1706.05350, 2017.

  • [13]↑

    A. G. Ganieand and S. Dadvandipour, “Identification of online harassment using ensemble fine-tuned pre-trained Bert,” Pollack Period., vol. 17, no. 3, pp. 13–18, 2022.

    • Search Google Scholar
    • Export Citation
  • [14]↑

    O. Dhifallah and Y. Lu, “On the inherent regularization effects of noise injection during training,” in Proceedings of the International Conference on Machine Learning, Virtual Event, July 18–24, 2021, pp. 2665–2675.

    • Search Google Scholar
    • Export Citation
  • [15]↑

    K. Zhang, Y. Li, W. Zuo, L. Zhang, L. V. Gool, and R. Timofte, “Plug-and-play image restoration with deep denoiser prior,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 10, pp. 6360–6376, 2021.

    • Search Google Scholar
    • Export Citation
  • [16]↑

    A. Pretorius, H. Kamper, and S. Kroon, “On the expected behaviour of noise regularised deep neural networks as Gaussian processes,” Pattern Recognit. Lett., vol. 138, pp. 75–81, 2020.

    • Search Google Scholar
    • Export Citation
  • [17]↑

    M. Qiao, S. Yan, X. Tang, and C. Xu, “Deep convolutional and LSTM recurrent neural networks for rolling bearing fault diagnosis under strong noises and variable loads,” IEEE Access, vol. 8, pp. 66257–66269, 2020.

    • Search Google Scholar
    • Export Citation
  • [18]↑

    A. A. Abdelhamid, E. S. M. El-Kenawy, B. Alotaibi, G. M. Amer, M. Y. Abdelkader, A. Ibrahim, and M. M. Eid, “Robust speech emotion recognition using CNN+ LSTM based on stochastic fractal search optimization algorithm,” IEEE Access, vol. 10, pp. 49265–49284, 2022.

    • Search Google Scholar
    • Export Citation
  • [19]↑

    G. Kovács, N. Yussupova, and D. Rizvanov, “Resource management simulation using multi-agent approach and semantic constraints,” Pollack Period., vol. 12, no. 1, pp. 45–58, 2017.

    • Search Google Scholar
    • Export Citation
  • [1]

    Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, pp. 436–444, 2015.

  • [2]

    C. M. Bishop, Pattern Recognition and Machine Learning. Springer, 2006.

  • [3]

    S. Ruder, “An overview of gradient descent optimization algorithms,” arXiv:1609.04747, 2016.

  • [4]

    I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. MIT Press, 2016.

  • [5]

    A. Neelakantan, L. Vilnis, Q. V. Le, I. Sutskever, L. Kaiser, K. Kurach, and J. Martens, “Adding gradient noise improves learning for very deep networks,” arXiv:1511.06807, 2015.

    • Search Google Scholar
    • Export Citation
  • [6]

    A. Graves and J. Schmidhuber, “Framewise phoneme classification with bidirectional LSTM and other neural network architectures,” Neural Netw., vol. 18, nos 5–6, pp. 602–610, 2005.

    • Search Google Scholar
    • Export Citation
  • [7]

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Comput., vol. 9, no. 8, pp. 1735–1780, 1997.

  • [8]

    P. Vincent, H. Larochelle, Y. Bengio, and P. A. Manzagol, “Extracting and composing robust features with denoising autoencoders,” in Proceedings of the 25th International Conference on Machine Learning, Helsinki, Finland, July 5–9, 2008, pp. 1096–1103.

    • Search Google Scholar
    • Export Citation
  • [9]

    Q. Zheng, M. Yang, J. Yang, Q. Zhang, and X. Zhang, “Improvement of generalization ability of deep CNN via implicit regularization in two-stage training process,” IEEE Access, vol. 6, pp. 15844–15869, 2018.

    • Search Google Scholar
    • Export Citation
  • [10]

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: a simple way to prevent neural networks from overfitting,” J. Machine Learn. Res., vol. 15, no. 1, pp. 1929–1958, 2014.

    • Search Google Scholar
    • Export Citation
  • [11]

    S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proceedings of the 32nd International Conference on International Conference on Machine Learning, Lille, France, July 6–11, 2015, pp. 448–456.

    • Search Google Scholar
    • Export Citation
  • [12]

    T. van Laarhoven, “L2 regularization versus batch and weight normalization,” arXiv:1706.05350, 2017.

  • [13]

    A. G. Ganieand and S. Dadvandipour, “Identification of online harassment using ensemble fine-tuned pre-trained Bert,” Pollack Period., vol. 17, no. 3, pp. 13–18, 2022.

    • Search Google Scholar
    • Export Citation
  • [14]

    O. Dhifallah and Y. Lu, “On the inherent regularization effects of noise injection during training,” in Proceedings of the International Conference on Machine Learning, Virtual Event, July 18–24, 2021, pp. 2665–2675.

    • Search Google Scholar
    • Export Citation
  • [15]

    K. Zhang, Y. Li, W. Zuo, L. Zhang, L. V. Gool, and R. Timofte, “Plug-and-play image restoration with deep denoiser prior,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 10, pp. 6360–6376, 2021.

    • Search Google Scholar
    • Export Citation
  • [16]

    A. Pretorius, H. Kamper, and S. Kroon, “On the expected behaviour of noise regularised deep neural networks as Gaussian processes,” Pattern Recognit. Lett., vol. 138, pp. 75–81, 2020.

    • Search Google Scholar
    • Export Citation
  • [17]

    M. Qiao, S. Yan, X. Tang, and C. Xu, “Deep convolutional and LSTM recurrent neural networks for rolling bearing fault diagnosis under strong noises and variable loads,” IEEE Access, vol. 8, pp. 66257–66269, 2020.

    • Search Google Scholar
    • Export Citation
  • [18]

    A. A. Abdelhamid, E. S. M. El-Kenawy, B. Alotaibi, G. M. Amer, M. Y. Abdelkader, A. Ibrahim, and M. M. Eid, “Robust speech emotion recognition using CNN+ LSTM based on stochastic fractal search optimization algorithm,” IEEE Access, vol. 10, pp. 49265–49284, 2022.

    • Search Google Scholar
    • Export Citation
  • [19]

    G. Kovács, N. Yussupova, and D. Rizvanov, “Resource management simulation using multi-agent approach and semantic constraints,” Pollack Period., vol. 12, no. 1, pp. 45–58, 2017.

    • Search Google Scholar
    • Export Citation
  • Collapse
  • Expand
Cover Pollack Periodica
Pollack Periodica
Print ISSN:
1788-1994
Online ISSN:
1788-3911

Senior editors

Editor(s)-in-Chief: Iványi, Amália

Editor(s)-in-Chief: Iványi, Péter

 

Scientific Secretary

Miklós M. Iványi

Editorial Board

  • Bálint Bachmann (Institute of Architecture, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Jeno Balogh (Department of Civil Engineering Technology, Metropolitan State University of Denver, Denver, Colorado, USA)
  • Radu Bancila (Department of Geotechnical Engineering and Terrestrial Communications Ways, Faculty of Civil Engineering and Architecture, “Politehnica” University Timisoara, Romania)
  • Charalambos C. Baniotopolous (Department of Civil Engineering, Chair of Sustainable Energy Systems, Director of Resilience Centre, School of Engineering, University of Birmingham, U.K.)
  • Oszkar Biro (Graz University of Technology, Institute of Fundamentals and Theory in Electrical Engineering, Austria)
  • Ágnes Borsos (Institute of Architecture, Department of Interior, Applied and Creative Design, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Matteo Bruggi (Dipartimento di Ingegneria Civile e Ambientale, Politecnico di Milano, Italy)
  • Petra Bujňáková (Department of Structures and Bridges, Faculty of Civil Engineering, University of Žilina, Slovakia)
  • Anikó Borbála Csébfalvi (Department of Civil Engineering, Institute of Smart Technology and Engineering, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Mirjana S. Devetaković (Faculty of Architecture, University of Belgrade, Serbia)
  • Szabolcs Fischer (Department of Transport Infrastructure and Water Resources Engineering, Faculty of Architerture, Civil Engineering and Transport Sciences Széchenyi István University, Győr, Hungary)
  • Radomir Folic (Department of Civil Engineering, Faculty of Technical Sciences, University of Novi Sad Serbia)
  • Jana Frankovská (Department of Geotechnics, Faculty of Civil Engineering, Slovak University of Technology in Bratislava, Slovakia)
  • János Gyergyák (Department of Architecture and Urban Planning, Institute of Architecture, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Kay Hameyer (Chair in Electromagnetic Energy Conversion, Institute of Electrical Machines, Faculty of Electrical Engineering and Information Technology, RWTH Aachen University, Germany)
  • Elena Helerea (Dept. of Electrical Engineering and Applied Physics, Faculty of Electrical Engineering and Computer Science, Transilvania University of Brasov, Romania)
  • Ákos Hutter (Department of Architecture and Urban Planning, Institute of Architecture, Faculty of Engineering and Information Technolgy, University of Pécs, Hungary)
  • Károly Jármai (Institute of Energy and Chemical Machinery, Faculty of Mechanical Engineering and Informatics, University of Miskolc, Hungary)
  • Teuta Jashari-Kajtazi (Department of Architecture, Faculty of Civil Engineering and Architecture, University of Prishtina, Kosovo)
  • Róbert Kersner (Department of Technical Informatics, Institute of Information and Electrical Technology, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Rita Kiss  (Biomechanical Cooperation Center, Faculty of Mechanical Engineering, Budapest University of Technology and Economics, Budapest, Hungary)
  • István Kistelegdi  (Department of Building Structures and Energy Design, Institute of Architecture, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Stanislav Kmeť (President of University Science Park TECHNICOM, Technical University of Kosice, Slovakia)
  • Imre Kocsis  (Department of Basic Engineering Research, Faculty of Engineering, University of Debrecen, Hungary)
  • László T. Kóczy (Department of Information Sciences, Faculty of Mechanical Engineering, Informatics and Electrical Engineering, University of Győr, Hungary)
  • Dražan Kozak (Faculty of Mechanical Engineering, Josip Juraj Strossmayer University of Osijek, Croatia)
  • György L. Kovács (Department of Technical Informatics, Institute of Information and Electrical Technology, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Balázs Géza Kövesdi (Department of Structural Engineering, Faculty of Civil Engineering, Budapest University of Engineering and Economics, Budapest, Hungary)
  • Tomáš Krejčí (Department of Mechanics, Faculty of Civil Engineering, Czech Technical University in Prague, Czech Republic)
  • Jaroslav Kruis (Department of Mechanics, Faculty of Civil Engineering, Czech Technical University in Prague, Czech Republic)
  • Miklós Kuczmann (Department of Automations, Faculty of Mechanical Engineering, Informatics and Electrical Engineering, Széchenyi István University, Győr, Hungary)
  • Tibor Kukai (Department of Engineering Studies, Institute of Smart Technology and Engineering, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Maria Jesus Lamela-Rey (Departamento de Construcción e Ingeniería de Fabricación, University of Oviedo, Spain)
  • János Lógó  (Department of Structural Mechanics, Faculty of Civil Engineering, Budapest University of Technology and Economics, Hungary)
  • Carmen Mihaela Lungoci (Faculty of Electrical Engineering and Computer Science, Universitatea Transilvania Brasov, Romania)
  • Frédéric Magoulés (Department of Mathematics and Informatics for Complex Systems, Centrale Supélec, Université Paris Saclay, France)
  • Gabriella Medvegy (Department of Interior, Applied and Creative Design, Institute of Architecture, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Tamás Molnár (Department of Visual Studies, Institute of Architecture, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Ferenc Orbán (Department of Mechanical Engineering, Institute of Smart Technology and Engineering, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Zoltán Orbán (Department of Civil Engineering, Institute of Smart Technology and Engineering, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Dmitrii Rachinskii (Department of Mathematical Sciences, The University of Texas at Dallas, Texas, USA)
  • Chro Radha (Chro Ali Hamaradha) (Sulaimani Polytechnic University, Technical College of Engineering, Department of City Planning, Kurdistan Region, Iraq)
  • Maurizio Repetto (Department of Energy “Galileo Ferraris”, Politecnico di Torino, Italy)
  • Zoltán Sári (Department of Technical Informatics, Institute of Information and Electrical Technology, Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Grzegorz Sierpiński (Department of Transport Systems and Traffic Engineering, Faculty of Transport, Silesian University of Technology, Katowice, Poland)
  • Zoltán Siménfalvi (Institute of Energy and Chemical Machinery, Faculty of Mechanical Engineering and Informatics, University of Miskolc, Hungary)
  • Andrej Šoltész (Department of Hydrology, Faculty of Civil Engineering, Slovak University of Technology in Bratislava, Slovakia)
  • Zsolt Szabó (Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Hungary)
  • Mykola Sysyn (Chair of Planning and Design of Railway Infrastructure, Institute of Railway Systems and Public Transport, Technical University of Dresden, Germany)
  • András Timár (Faculty of Engineering and Information Technology, University of Pécs, Hungary)
  • Barry H. V. Topping (Heriot-Watt University, UK, Faculty of Engineering and Information Technology, University of Pécs, Hungary)

POLLACK PERIODICA
Pollack Mihály Faculty of Engineering
Institute: University of Pécs
Address: Boszorkány utca 2. H–7624 Pécs, Hungary
Phone/Fax: (36 72) 503 650

E-mail: peter.ivanyi@mik.pte.hu 

or amalia.ivanyi@mik.pte.hu

Indexing and Abstracting Services:

  • SCOPUS
  • CABELLS Journalytics

 

2024  
Scopus  
CiteScore  
CiteScore rank  
SNIP  
Scimago  
SJR index 0.385
SJR Q rank Q3

2023  
Scopus  
CiteScore 1.5
CiteScore rank Q3 (Civil and Structural Engineering)
SNIP 0.849
Scimago  
SJR index 0.288
SJR Q rank Q3

Pollack Periodica
Publication Model Hybrid
Submission Fee none
Article Processing Charge 900 EUR/article
Printed Color Illustrations 40 EUR (or 10 000 HUF) + VAT / piece
Regional discounts on country of the funding agency World Bank Lower-middle-income economies: 50%
World Bank Low-income economies: 100%
Further Discounts Editorial Board / Advisory Board members: 50%
Corresponding authors, affiliated to an EISZ member institution subscribing to the journal package of Akadémiai Kiadó: 100%
Subscription fee 2025 Online subsscription: 381 EUR / 420 USD
Print + online subscription: 456 EUR / 520 USD
Subscription Information Online subscribers are entitled access to all back issues published by Akadémiai Kiadó for each title for the duration of the subscription, as well as Online First content for the subscribed content.
Purchase per Title Individual articles are sold on the displayed price.

 

2023  
Scopus  
CiteScore 1.5
CiteScore rank Q3 (Civil and Structural Engineering)
SNIP 0.849
Scimago  
SJR index 0.288
SJR Q rank Q3

Monthly Content Usage

Abstract Views Full Text Views PDF Downloads
Nov 2024 0 107 25
Dec 2024 0 387 319
Jan 2025 0 271 149
Feb 2025 0 186 36
Mar 2025 0 149 32
Apr 2025 0 80 21
May 2025 0 0 0