Books
Sorry, no publications matched your criteria.
Book Chapters
1.
Tsakalos K, Sirakoulis G Ch, Adamatzky A
Unsupervised Learning Approach Using Reinforcement Techniques on Bio-inspired Topologies Book Chapter
In: Handbook of Unconventional Computing: VOLUME 1: Theory, vol. 1, Chapter 17, pp. 507–533, World Scientific, 2021.
@inbook{tsakalos2021unsupervised,
title = {Unsupervised Learning Approach Using Reinforcement Techniques on Bio-inspired Topologies},
author = {Karolos-Alexandros Tsakalos and Georgios Ch. Sirakoulis and Andrew Adamatzky},
url = {https://www.worldscientific.com/doi/10.1142/9789811235726_0017},
doi = {doi.org/10.1142/9789811235726_0017},
year = {2021},
date = {2021-10-19},
urldate = {2021-10-19},
booktitle = {Handbook of Unconventional Computing: VOLUME 1: Theory},
volume = {1},
pages = {507--533},
publisher = {World Scientific},
chapter = {17},
abstract = {Modeling complex bio-inspired networks is widely used in the research field of emerging computing, which promises rapid growth in the field of computer science. This work deals with bio-inspired molecular networks which have been studied through neuromorphic computing. This molecular-based structure is adapted to create a complex recurrent neuromorphic network that consists of neurons integrated with the simple Izhikevich neuromorphic model. Therefore molecular atoms are considered as neurons and chemical edges as synapses. More specifically, the molecular-based structure of Verotoxin-1 molecule has been extensively studied. Two Reinforcement excitation techniques inspired from Cellular Automata studies, namely, Game-of-Life (GoL)-rule and Majority-rule, are employed to control the stimulation of each neuron depending its neighbourhood activity. In this work, two different CA-inspired unsupervised learning methods along with the neuro-inspired Hebbian learning have also utilized the local activity to apply self-organization and update the recurrent synaptic weights to highlight complex neuromorphic clusters that are integrated into the existing molecular structure. Finally, by applying the proposed reinforcement excitation techniques along with the unsupervised learning, we investigate the potential of spatio-temporal signals classification through the proposed framework based on the molecular structure. The obtained results showed us this framework ability of of distinguishing high-dimensional signals; in this sense, we further discuss about how these learning approach along with molecular-based structures can be utilized to learn and help us with different complex tasks in a wide range of applications such as the classification of multi-dimensional signals.},
keywords = {},
pubstate = {published},
tppubtype = {inbook}
}
Modeling complex bio-inspired networks is widely used in the research field of emerging computing, which promises rapid growth in the field of computer science. This work deals with bio-inspired molecular networks which have been studied through neuromorphic computing. This molecular-based structure is adapted to create a complex recurrent neuromorphic network that consists of neurons integrated with the simple Izhikevich neuromorphic model. Therefore molecular atoms are considered as neurons and chemical edges as synapses. More specifically, the molecular-based structure of Verotoxin-1 molecule has been extensively studied. Two Reinforcement excitation techniques inspired from Cellular Automata studies, namely, Game-of-Life (GoL)-rule and Majority-rule, are employed to control the stimulation of each neuron depending its neighbourhood activity. In this work, two different CA-inspired unsupervised learning methods along with the neuro-inspired Hebbian learning have also utilized the local activity to apply self-organization and update the recurrent synaptic weights to highlight complex neuromorphic clusters that are integrated into the existing molecular structure. Finally, by applying the proposed reinforcement excitation techniques along with the unsupervised learning, we investigate the potential of spatio-temporal signals classification through the proposed framework based on the molecular structure. The obtained results showed us this framework ability of of distinguishing high-dimensional signals; in this sense, we further discuss about how these learning approach along with molecular-based structures can be utilized to learn and help us with different complex tasks in a wide range of applications such as the classification of multi-dimensional signals.
Editorials
Sorry, no publications matched your criteria.
Journals
4.
Tsakalos K, Sirakoulis G Ch, Adamatzky A, Smith J
Protein Structured Reservoir computing for Spike-based Pattern Recognition Journal Article
In: IEEE Transactions on Parallel and Distributed Systems, vol. 33, no. 2, pp. 322–331, 2022.
@article{tsakalos2021protein,
title = {Protein Structured Reservoir computing for Spike-based Pattern Recognition},
author = {Karolos-Alexandros Tsakalos and Georgios Ch. Sirakoulis and Andy Adamatzky and Jim Smith},
url = {https://www.computer.org/csdl/journal/td/2022/02/09387584/1smD8QlWRji},
doi = {10.1109/TPDS.2021.3068826},
year = {2022},
date = {2022-02-01},
urldate = {2022-02-01},
journal = {IEEE Transactions on Parallel and Distributed Systems},
volume = {33},
number = {2},
pages = {322--331},
publisher = {IEEE},
abstract = {Nowadays we witness a miniaturisation trend in the semiconductor industry backed up by groundbreaking discoveries and designs in nanoscale characterisation and fabrication. To facilitate the trend and produce ever smaller, faster and cheaper computing devices, the size of nanoelectronic devices is now reaching the scale of atoms or molecules - a technical goal undoubtedly demanding for novel devices. Following the trend, we explore an unconventional route of implementing reservoir computing on a single protein molecule and introduce neuromorphic connectivity with a small-world networking property. We have chosen Izhikevich spiking neurons as elementary processors, corresponding to the atoms of verotoxin protein, and its molecule as a `hardware' architecture of the communication networks connecting the processors. We apply on a single readout layer, various training methods in a supervised fashion to investigate whether the molecular structured Reservoir Computing (RC) system is capable to deal with machine learning benchmarks. We start with the Remote Supervised Method, based on Spike-Timing-Dependent-Plasticity, and carry on with linear regression and scaled conjugate gradient back-propagation training methods. The RC network is evaluated as a proof-of-concept on the handwritten digit images from the standard MNIST and the extended MNIST datasets and demonstrates acceptable classification accuracies in comparison with other similar approaches},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Nowadays we witness a miniaturisation trend in the semiconductor industry backed up by groundbreaking discoveries and designs in nanoscale characterisation and fabrication. To facilitate the trend and produce ever smaller, faster and cheaper computing devices, the size of nanoelectronic devices is now reaching the scale of atoms or molecules - a technical goal undoubtedly demanding for novel devices. Following the trend, we explore an unconventional route of implementing reservoir computing on a single protein molecule and introduce neuromorphic connectivity with a small-world networking property. We have chosen Izhikevich spiking neurons as elementary processors, corresponding to the atoms of verotoxin protein, and its molecule as a `hardware' architecture of the communication networks connecting the processors. We apply on a single readout layer, various training methods in a supervised fashion to investigate whether the molecular structured Reservoir Computing (RC) system is capable to deal with machine learning benchmarks. We start with the Remote Supervised Method, based on Spike-Timing-Dependent-Plasticity, and carry on with linear regression and scaled conjugate gradient back-propagation training methods. The RC network is evaluated as a proof-of-concept on the handwritten digit images from the standard MNIST and the extended MNIST datasets and demonstrates acceptable classification accuracies in comparison with other similar approaches
3.
Tsakalos K, Dragkola P, Karamani R, Tsompanas M, Provata A, Dimitrakis P, Adamatzky A I, Sirakoulis G C
Chimera States in Neuro-Inspired Area-Efficient Asynchronous Cellular Automata Networks Journal Article
In: IEEE Transactions on Circuits and Systems I: Regular Papers, 2022.
@article{tsakalos2022chimera,
title = {Chimera States in Neuro-Inspired Area-Efficient Asynchronous Cellular Automata Networks},
author = {Karolos-Alexandros Tsakalos and Paraskevi Dragkola and Rafailia-Eleni Karamani and Michail-Antisthenis Tsompanas and Astero Provata and Panagiotis Dimitrakis and Andrew I Adamatzky and Georgios Ch Sirakoulis},
year = {2022},
date = {2022-01-01},
journal = {IEEE Transactions on Circuits and Systems I: Regular Papers},
publisher = {IEEE},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
2.
Karamani R, Fyrigos I, Tsakalos K, Ntinas V, Tsompanas M, Sirakoulis G Ch
Memristive learning cellular automata for edge detection Journal Article
In: Chaos, Solitons & Fractals, vol. 145, pp. 110700, 2021.
@article{karamani2021memristive,
title = {Memristive learning cellular automata for edge detection},
author = {Rafailia-Eleni Karamani and Iosif-Angelos Fyrigos and Karolos-Alexandros Tsakalos and Vasileios Ntinas and Michail-Antisthenis Tsompanas and Georgios Ch. Sirakoulis},
url = {https://www.sciencedirect.com/science/article/abs/pii/S0960077921000539},
doi = {doi.org/10.1016/j.chaos.2021.110700},
year = {2021},
date = {2021-02-25},
urldate = {2021-02-25},
journal = {Chaos, Solitons \& Fractals},
volume = {145},
pages = {110700},
publisher = {Elsevier},
abstract = {Memristors have been utilized as an unconventional computational substrate and gained interest as a medium to implement neuromorphic computations. A mathematical model that also proved its potential is Learning Cellular Automata, that is an amalgam of Cellular Automata and Learning Automata. The realization of the common characteristics of memristive circuits and Learning Cellular Automata can only lead to their combination. Namely, both manage to blend storage and processing capabilities in their basic entity. This study involves the definition of memristive circuits that realize the computing behavior of Learning Cellular Automata. An example of this methodology is provided with the description of the implementation of edge detection for image processing.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Memristors have been utilized as an unconventional computational substrate and gained interest as a medium to implement neuromorphic computations. A mathematical model that also proved its potential is Learning Cellular Automata, that is an amalgam of Cellular Automata and Learning Automata. The realization of the common characteristics of memristive circuits and Learning Cellular Automata can only lead to their combination. Namely, both manage to blend storage and processing capabilities in their basic entity. This study involves the definition of memristive circuits that realize the computing behavior of Learning Cellular Automata. An example of this methodology is provided with the description of the implementation of edge detection for image processing.
1.
Floros T, Tsakalos K, Dourvas N, Tsompanas M, Sirakoulis G Ch
Unconventional Bio-Inspired Model for Design of Logic Gates Journal Article
In: International Journal of Unconventional Computing, vol. 15, no. 3, pp. 141–156, 2020.
@article{floros2020unconventional,
title = {Unconventional Bio-Inspired Model for Design of Logic Gates},
author = {Theofanis Floros and Karolos-Alexandros Tsakalos and Nikolaos Dourvas and Michail-Antisthenis Tsompanas and Georgios Ch. Sirakoulis},
url = {https://www.oldcitypublishing.com/journals/ijuc-home/ijuc-issue-contents/ijuc-volume-15-number-3-2020/ijuc-15-3-p-141-156/},
year = {2020},
date = {2020-08-01},
urldate = {2020-01-01},
journal = {International Journal of Unconventional Computing},
volume = {15},
number = {3},
pages = {141--156},
publisher = {Old City Publishing},
abstract = {During the last years, a well studied biological substrate, namely Physarum polycephalum, has been proven efficient on finding appropriate and efficient solutions in hard to solve complex mathematical problems. The plasmodium of P. polycephalum is a single-cell that serves as a prosperous bio-computational example. Consequently, it has been successfully utilized in the past to solve a variety of path problems in graphs and combinatorial problems. In this work, this interesting behaviour is mimicked by a robust unconventional computational model, drawing inspiration from the notion of Cellular and Learning Automata. Namely, we employ principles of Cellular Automata (CAs) enriched with learning capabilities to develop a robust computational model, able of modelling appropriately the aforementioned biological substrate and, thus, capturing its computational capabilities. CAs are very efficient in modelling biological systems and solving scientific problems, owing to their ability of incarnating essential properties of a system where global behaviour arises as an effect of simple components, interacting locally. The resulting computational tool, after combining CAs with learning capabilities, should be appropriate for modelling the behaviour of living organisms. Thus, the inherent abilities and computational characteristics of the proposed bio-inspired model are stressed towards the experimental verification of Physarum’s ability to model Logic Gates, while trying to find minimal paths in properly configured mazes with food sources. The presented simulation results for various Logic Gates are found in good agreement, both qualitatively and quantitatively, with the corresponding experimental results, proving the efficacy of this unconventional bio-inspired model and providing useful insights for its enhanced usage in various computing applications.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
During the last years, a well studied biological substrate, namely Physarum polycephalum, has been proven efficient on finding appropriate and efficient solutions in hard to solve complex mathematical problems. The plasmodium of P. polycephalum is a single-cell that serves as a prosperous bio-computational example. Consequently, it has been successfully utilized in the past to solve a variety of path problems in graphs and combinatorial problems. In this work, this interesting behaviour is mimicked by a robust unconventional computational model, drawing inspiration from the notion of Cellular and Learning Automata. Namely, we employ principles of Cellular Automata (CAs) enriched with learning capabilities to develop a robust computational model, able of modelling appropriately the aforementioned biological substrate and, thus, capturing its computational capabilities. CAs are very efficient in modelling biological systems and solving scientific problems, owing to their ability of incarnating essential properties of a system where global behaviour arises as an effect of simple components, interacting locally. The resulting computational tool, after combining CAs with learning capabilities, should be appropriate for modelling the behaviour of living organisms. Thus, the inherent abilities and computational characteristics of the proposed bio-inspired model are stressed towards the experimental verification of Physarum’s ability to model Logic Gates, while trying to find minimal paths in properly configured mazes with food sources. The presented simulation results for various Logic Gates are found in good agreement, both qualitatively and quantitatively, with the corresponding experimental results, proving the efficacy of this unconventional bio-inspired model and providing useful insights for its enhanced usage in various computing applications.
Conferences
1.
Tsakalos , Karolos-Alexandros , Ntinas V, Karamani R, Fyrigos I, Chatzinikolaou T P, Vasileiadis N, Dimitrakis P, Provata A, Sirakoulis G Ch
Emergence of Chimera States with Re-programmable Memristor Crossbar Arrays Proceedings Article
In: 2021 IEEE International Symposium on Circuits and Systems (ISCAS), pp. 1–5, IEEE 2021.
@inproceedings{tsakalos2021emergence,
title = {Emergence of Chimera States with Re-programmable Memristor Crossbar Arrays},
author = {Tsakalos and Karolos-Alexandros and Vasileios Ntinas and Rafailia-Eleni Karamani and Iosif-Angelos Fyrigos and Theodoros Panagiotis Chatzinikolaou and Nikolaos Vasileiadis and Panagiotis Dimitrakis and Astero Provata and Georgios Ch. Sirakoulis},
url = {https://ieeexplore.ieee.org/document/9401669},
doi = {10.1109/ISCAS51556.2021.9401669},
year = {2021},
date = {2021-04-27},
urldate = {2021-01-01},
booktitle = {2021 IEEE International Symposium on Circuits and Systems (ISCAS)},
pages = {1--5},
organization = {IEEE},
abstract = {The time series of the brain are usually characterized by the co-existence of synchronized and desynchronized behaviors. This kind of behavior is related to normal and disorderly functions of the brain. One of the suggested mechanisms to understand thoroughly this behavior are chimera states, which are characterized by the coincidence of coherent and incoherent dynamics that can be exploited through networks of symmetrically coupled identical oscillators. In this work, ring-based networks of Chua's circuits, the simplest electronic oscillators that perform chaotic and well-known bifurcation phenomena, have been extensively studied in memristive crossbars (Xbar), revealing various collective spatio-temporal behaviors, such as chimera states. With respect to different Xbar connectivities and via SPICE-level circuit simulations, the proposed Xbar system proves its efficacy to reproduce spatio-temporal patterns spanning from complete synchronization and chimera states up to fully chaotic states.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The time series of the brain are usually characterized by the co-existence of synchronized and desynchronized behaviors. This kind of behavior is related to normal and disorderly functions of the brain. One of the suggested mechanisms to understand thoroughly this behavior are chimera states, which are characterized by the coincidence of coherent and incoherent dynamics that can be exploited through networks of symmetrically coupled identical oscillators. In this work, ring-based networks of Chua's circuits, the simplest electronic oscillators that perform chaotic and well-known bifurcation phenomena, have been extensively studied in memristive crossbars (Xbar), revealing various collective spatio-temporal behaviors, such as chimera states. With respect to different Xbar connectivities and via SPICE-level circuit simulations, the proposed Xbar system proves its efficacy to reproduce spatio-temporal patterns spanning from complete synchronization and chimera states up to fully chaotic states.