Off the Mainstream: Advances in Neural Networks and Machine Learning for Pattern Recognition

• Published in: Neural processing letters Vol. 48; no. 2; pp. 643 - 648
• Main Authors: Trentin, Edmondo, Schwenker, Friedhelm, El Gayar, Neamat, Abbas, Hazem M.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.10.2018; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Artificial neural networks; Bayesian analysis; Bioinformatics; Complex Systems; Computational Intelligence; Computer Science; Deep learning; Fields (mathematics); Graphs; Machine learning; Neural networks; Optical character recognition; Pattern recognition; Pipelining (computers); Publishing; Sequences; Taxonomy; Time series; Trends
• ISSN: 1370-4621; 1573-773X
• DOI: 10.1007/s11063-018-9830-8

The first paper in this grouping, namely A Temporal Dependency based Multi-modal Active Learning Approach for Audiovisual Event Detection (an extension of paper [23], presented at ANNPR2016) by Patrick Thiam, Sascha Meudt, Günther Palm and Friedhelm Schwenker, proposes a general transfer learning mechanism [7] which relies on a novel strategy for the automatic labeling of unlabeled samples in a partially supervised setup [19]. In the next paper, titled Dynamic Hybrid Random Fields for the Probabilistic Graphical Modeling of Sequential Data: Definitions, Algorithms, and an Application to Bioinformatics (extended version of the ANNPR2016 paper [4]), Marco Bongini, Antonino Freno, Vincenzo Laveglia and Edmondo Trentin formalize the notion of a dynamic extension to sequences of a probabilistic graphical model (the hybrid random field [5, 25]) that subsumes Bayesian networks and Markov random fields [11], presenting parameter and structure learning algorithms and an application to the prediction of the disulfide bonding state of cysteines [28]. [...]Mohamed A. Radwan, Mahmoud Khalil and Hazem Abbas, in their paper Neural Networks Pipeline for Offline Machine Printed Arabic OCR (an extended version of their ANNPR2016 paper [15]) present a viable solution to the problem of Arabic optical characters recognition relying on a pipeline of three separate neural modules taking responsibility, respectively, for: (i) font size normalization, (ii) word segmentation into characters, and (iii) character recognition via a convolutional neural network. [...]Witali Aswolinskiy, Felix Reinhart and Jochen Steil, in their paper Time Series Classification in Reservoir- and Model-Space (extended version of their ANNPR2016 paper [2]), present a large-scale experimental investigation of the prediction of both univariate and multivariate time series relying on reservoir computing [13].