YoonSik Shim - Academia.edu (original) (raw)

Papers by YoonSik Shim

Research paper thumbnail of Examples of STDP weight maps from different feature selection schemes when <i>N</i><sub><i>E</i></sub> = 5

<p>The weight maps for the ensemble WTA neurons which represent the digit 1 after learning ... more <p>The weight maps for the ensemble WTA neurons which represent the digit 1 after learning are shown.</p

Research paper thumbnail of SEM-ITDP ensemble network architecture

<p>The STDP connections, which projects from the selected input neurons to each WTA circuit... more <p>The STDP connections, which projects from the selected input neurons to each WTA circuit, together with the WTA circuits constitute the SEM ensemble. The ITDP connections have the same connectivity as the logical ITDP model. All of the ensemble, gating and final output networks use the same SEM circuit model.</p

Research paper thumbnail of The standard MoE architecture

<p>The outputs (classifications) from the classifier networks are fed into an output unit w... more <p>The outputs (classifications) from the classifier networks are fed into an output unit which combines them according to some simple rule. The gating network weights the individual classifier outputs before they enter the final output unit, and thus guides learning of the overall combined classification. The classifiers and gating networks receive the same input data. See text for further details.</p

Research paper thumbnail of Illustrative images for controlled feature assignment for SEM ensemble networks

<p>White regions indicate available pixels (active region) as defined by preprocessing, and... more <p>White regions indicate available pixels (active region) as defined by preprocessing, and the Gaussian means for the normal Gaussian selection scheme are evenly placed inside such regions by random placement procedure (See Methods for details of the actual Gaussian mean placement). The number of stretched Gaussian features used increases linearly with ensemble size (see <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1005137#sec015&quot; target="_blank">Methods</a> for details). The diameters of red circles and ovals roughly represent the full width at a tenth of maximum (FWTM) for each principal direction (the length of an oval is shown far shorter than it actual is for the sake of visualization—long ovals are used to ensure they form roughly uniform bars in the region of available pixels). In all cases, exactly 1/4 of pixels from the available (white) region are stochastically selected (without replacement) for each ensemble network according to each distribution function.</p

Research paper thumbnail of Examples of ensemble behaviours (<i>N</i><sub><i>E</i></sub> = 9) for different gating network performances ((A) better than, (B) similar to, (C) worse than the ensemble average)

<p>All the ensemble and the gating WTAs used random feature selection. The colors represent... more <p>All the ensemble and the gating WTAs used random feature selection. The colors represent the NCEs of the final network (red), the gating network (blue), the ensemble networks (grey) and their average (black). Vertical lines indicate the time span of the total data presentation, where input data are sequentially presented for multiple rounds in order to see long term convergence. The NCE value at time <i>t</i> is calculated by counting the class-dependent spikes within the past finite time window of [<i>T</i><sub><i>p</i></sub>, <i>t</i>] (<i>T</i><sub><i>p</i></sub> < <i>t</i>). In order to prevent a sudden change in the NCE plots due to the exclusion of the early system output (which are immature resulting in high NCE values) from the time window, <i>T</i><sub><i>p</i></sub> was dynamically changed for faster burn-out of those initial values as: <i>T</i><sub><i>p</i></sub> = <i>t</i>(1−<i>d</i>/4<i>D</i>) where <i>d</i> = <i>t</i> when <i>t</i> < 2<i>D</i> and <i>d</i> = 2<i>D</i> otherwise, <i>D</i> = 224sec is the duration of one round of dataset presentation. See Methods for details of the NCE calculations.</p

Research paper thumbnail of An example of the STDP weight maps of a SEM classifier after learning (A, B) and the time evolution of ITDP weights (C)

<p>Each weight map represents the presynaptic weight values that project to each of four WT... more <p>Each weight map represents the presynaptic weight values that project to each of four WTA neurons (which each fire dominantly for one of the classes). The grey area shows pixels disabled by preprocessing, and each colored pixel represent the difference of the weights from the two input neurons for the corresponding pixel (white pixels represent unselected features). So as to use all features, a quarter of pixels are evenly selected from the supersampled image in order to use all pixels of the original data.</p

Research paper thumbnail of Unsupervised Learning in an Ensemble of Spiking Neural Networks Mediated by ITDP - Fig 12

Unsupervised learning in an ensemble of spiking neural networks mediated by ITDP Article (Publish... more Unsupervised learning in an ensemble of spiking neural networks mediated by ITDP Article (Published Version) http://sro.sussex.ac.uk Shim, Yoonsik, Philippides, Andy, Staras, Kevin and Husbands, Phil (2016) Unsupervised learning in an ensemble of spiking neural networks mediated by ITDP. PLoS Computational Biology, 12 (10). e1005137.

Research paper thumbnail of Spike trains from the SEM ensemble network with <i>N</i><sub><i>E</i></sub> = 5 and random feature selection

<p>(Left) Plot shows the input neuron spikes from eight image presentations from different ... more <p>(Left) Plot shows the input neuron spikes from eight image presentations from different classes (digits) which are depicted in different colors (black: 0, red: 1, green: 2, blue: 4). (Right) Two graphs show the output spikes of ensemble, gating, and final WTA neurons before and after learning. The colors of the spikes represent which class is being presented as input. After learning the network outputs produce consistent firing patterns, each output spiking exclusively for a single class.</p

Research paper thumbnail of Examples of random Gaussian mean placements for different <i>N</i><sub><i>E</i></sub> from the manually designed initial points (black points)

<p>The red pixels represent the outer border of the active region of the image, and the yel... more <p>The red pixels represent the outer border of the active region of the image, and the yellow pixels represent a forbidden region which is 3 pixels thick. The jittered mean points were restricted to be placed inside the inner region (including the green pixels) which is surrounded by the inner border (green).</p

Research paper thumbnail of Training performances of the expanded STDP/ITDP networks (using random feature selection on the MNIST handwritten digits classification task as in earlier experiments)

<p>Each color represents, red: ITDP final WTA, green: STDP final WTA, blue: gating WTA, gre... more <p>Each color represents, red: ITDP final WTA, green: STDP final WTA, blue: gating WTA, grey/black: ensemble WTAs and their average. (A, B) An example of time courses of performances and the final performances from 50 repeated trials using unsupervised gating WTA. The individual trials were sorted by gating WTA performances in ascending order. (C, D) Simulations using the automatic selection of gating WTA. The vertical lines with arrowheads in C indicate where the switching of gating WTA occurs (see text for further details).</p

Research paper thumbnail of A voter and the voter ensemble network (<i>N</i><sub><i>C</i></sub> = 4)

<p>(Left) A voter and the predefined firing probabilities of each voter neuron for a set of... more <p>(Left) A voter and the predefined firing probabilities of each voter neuron for a set of virtual input samples <i>X</i> = {<i>x</i><sub>1</sub>, <i>x</i><sub>2</sub>, …, <i>x</i><sub><i>M</i></sub>}. (Right) The voter ensemble network. The weight represents the weight of connection from the <i>i</i>th neuron of the <i>j</i>th voter to the <i>k</i>th neuron of the final voter.</p

Research paper thumbnail of 순응적 날개를 가진 날갯짓 로봇 시뮬레이션을 위한 뉴로컨트롤의 진화연산 최적화

The Korean Society Of Computer And Information, Dec 1, 2019

Research paper thumbnail of 경로 제어가 가능한 가상생명체를 위한 2단계 진화 알고리즘

Research paper thumbnail of Learning Media on Mathematical Education based on Augmented Reality

KSII Transactions on Internet and Information Systems, 2021

Modern technology offers many ways to enhance teaching and learning that in turn promote the deve... more Modern technology offers many ways to enhance teaching and learning that in turn promote the development of tools for educational activities both inside and outside the classroom. Many educational programs using the augmented reality (AR) technology are being widely used to provide supplementary learning materials for students. This paper describes the potential and challenges of using GeoGebra AR in mathematical studies, whereby students can view 3D geometric objects for a better understanding of their structure, and verifies the feasibility of its use based on experimental results. The GeoGebra software can be used to draw geometric objects, and 3D geometric objects can be viewed using AR software or AR applications on mobile phones or computer tablets. These could provide some of the required materials for mathematical education at high schools or universities. The use of the GeoGebra application for education in Laos will be particularly discussed in this paper.

Research paper thumbnail of Evolutionary Robotics and Neuroscience

The Horizons of Evolutionary Robotics, 2014

Research paper thumbnail of Generating flying creatures using body-brain co-evolution

This paper describes a system that produces double-winged flying creatures using body-brain co-ev... more This paper describes a system that produces double-winged flying creatures using body-brain co-evolution without need of complex flapping flight aerodynamics. While artificial life techniques have been used to create a variety of virtual creatures, little work has explored flapping-winged creatures for the difficulty of genetic encoding problem of wings with limited geometric primitives as well as flapping-wing aerodynamics. Despite of the simplicity of system, our result shows aesthetical looking and organic flapping flight locomotions. The restricted list structure is used in genotype encoding for morphological symmetry of creatures and is more easily handled than other data structures. The creatures evolved by this system have two symmetric flapping wings consisting of continuous triangular patches and show various looking and locomotion such as wings of birds, butterflies and bats or even imaginary wings of a dragon and pterosaurs.

Research paper thumbnail of Evolutionary Optimization of Neurocontroller for Physically Simulated Compliant-Wing Ornithopter

Journal of the Korea Society of Computer and Information, 2019

Research paper thumbnail of Recent advances in evolutionary and bio-inspired adaptive robotics: Exploiting embodied dynamics

Applied Intelligence, 2021

This paper explores current developments in evolutionary and bio-inspired approaches to autonomou... more This paper explores current developments in evolutionary and bio-inspired approaches to autonomous robotics, concentrating on research from our group at the University of Sussex. These developments are discussed in the context of advances in the wider fields of adaptive and evolutionary approaches to AI and robotics, focusing on the exploitation of embodied dynamics to create behaviour. Four case studies highlight various aspects of such exploitation. The first exploits the dynamical properties of a physical electronic substrate, demonstrating for the first time how component-level analog electronic circuits can be evolved directly in hardware to act as robot controllers. The second develops novel, effective and highly parsimonious navigation methods inspired by the way insects exploit the embodied dynamics of innate behaviours. Combining biological experiments with robotic modeling, it is shown how rapid route learning can be achieved with the aid of navigation-specific visual info...

Research paper thumbnail of Evolving Flying Creatures with Path Following Behaviors

Artificial Life IX, 2004

We present a system which evolves physically simulated 3D flying creatures and their maneuvers. T... more We present a system which evolves physically simulated 3D flying creatures and their maneuvers. The creature is modelled as a number of articulated cylinders connected by triangular patagia in between. A creature's wing structure and its low-level controllers for straight flight are generated by an evolutionary algorithm. Then a feed-forward neural network is attached to the low-level controllers, and the connection weights of the network for a given trajectory are found by a genetic algonthm. We show that a control system sufficiently effective to allow aerial creatures to follow a complicated path can be achieved by two-step evolution process.

Research paper thumbnail of Embodied neuromechanical chaos through homeostatic regulation

Chaos: An Interdisciplinary Journal of Nonlinear Science, 2019

In this paper, we present detailed analyses of the dynamics of a number of embodied neuromechanic... more In this paper, we present detailed analyses of the dynamics of a number of embodied neuromechanical systems of a class that has been shown to efficiently exploit chaos in the development and learning of motor behaviors for bodies of arbitrary morphology. This class of systems has been successfully used in robotics, as well as to model biological systems. At the heart of these systems are neural central pattern generating (CPG) units connected to actuators which return proprioceptive information via an adaptive homeostatic mechanism. Detailed dynamical analyses of example systems, using high resolution largest Lyapunov exponent maps, demonstrate the existence of chaotic regimes within a particular region of parameter space, as well as the striking similarity of the maps for systems of varying size. Thanks to the homeostatic sensory mechanisms, any single CPG “views” the whole of the rest of the system as if it was another CPG in a two coupled system, allowing a scale invariant concep...

Research paper thumbnail of Examples of STDP weight maps from different feature selection schemes when <i>N</i><sub><i>E</i></sub> = 5

<p>The weight maps for the ensemble WTA neurons which represent the digit 1 after learning ... more <p>The weight maps for the ensemble WTA neurons which represent the digit 1 after learning are shown.</p

Research paper thumbnail of SEM-ITDP ensemble network architecture

<p>The STDP connections, which projects from the selected input neurons to each WTA circuit... more <p>The STDP connections, which projects from the selected input neurons to each WTA circuit, together with the WTA circuits constitute the SEM ensemble. The ITDP connections have the same connectivity as the logical ITDP model. All of the ensemble, gating and final output networks use the same SEM circuit model.</p

Research paper thumbnail of The standard MoE architecture

<p>The outputs (classifications) from the classifier networks are fed into an output unit w... more <p>The outputs (classifications) from the classifier networks are fed into an output unit which combines them according to some simple rule. The gating network weights the individual classifier outputs before they enter the final output unit, and thus guides learning of the overall combined classification. The classifiers and gating networks receive the same input data. See text for further details.</p

Research paper thumbnail of Illustrative images for controlled feature assignment for SEM ensemble networks

<p>White regions indicate available pixels (active region) as defined by preprocessing, and... more <p>White regions indicate available pixels (active region) as defined by preprocessing, and the Gaussian means for the normal Gaussian selection scheme are evenly placed inside such regions by random placement procedure (See Methods for details of the actual Gaussian mean placement). The number of stretched Gaussian features used increases linearly with ensemble size (see <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1005137#sec015&quot; target="_blank">Methods</a> for details). The diameters of red circles and ovals roughly represent the full width at a tenth of maximum (FWTM) for each principal direction (the length of an oval is shown far shorter than it actual is for the sake of visualization—long ovals are used to ensure they form roughly uniform bars in the region of available pixels). In all cases, exactly 1/4 of pixels from the available (white) region are stochastically selected (without replacement) for each ensemble network according to each distribution function.</p

Research paper thumbnail of Examples of ensemble behaviours (<i>N</i><sub><i>E</i></sub> = 9) for different gating network performances ((A) better than, (B) similar to, (C) worse than the ensemble average)

<p>All the ensemble and the gating WTAs used random feature selection. The colors represent... more <p>All the ensemble and the gating WTAs used random feature selection. The colors represent the NCEs of the final network (red), the gating network (blue), the ensemble networks (grey) and their average (black). Vertical lines indicate the time span of the total data presentation, where input data are sequentially presented for multiple rounds in order to see long term convergence. The NCE value at time <i>t</i> is calculated by counting the class-dependent spikes within the past finite time window of [<i>T</i><sub><i>p</i></sub>, <i>t</i>] (<i>T</i><sub><i>p</i></sub> < <i>t</i>). In order to prevent a sudden change in the NCE plots due to the exclusion of the early system output (which are immature resulting in high NCE values) from the time window, <i>T</i><sub><i>p</i></sub> was dynamically changed for faster burn-out of those initial values as: <i>T</i><sub><i>p</i></sub> = <i>t</i>(1−<i>d</i>/4<i>D</i>) where <i>d</i> = <i>t</i> when <i>t</i> < 2<i>D</i> and <i>d</i> = 2<i>D</i> otherwise, <i>D</i> = 224sec is the duration of one round of dataset presentation. See Methods for details of the NCE calculations.</p

Research paper thumbnail of An example of the STDP weight maps of a SEM classifier after learning (A, B) and the time evolution of ITDP weights (C)

<p>Each weight map represents the presynaptic weight values that project to each of four WT... more <p>Each weight map represents the presynaptic weight values that project to each of four WTA neurons (which each fire dominantly for one of the classes). The grey area shows pixels disabled by preprocessing, and each colored pixel represent the difference of the weights from the two input neurons for the corresponding pixel (white pixels represent unselected features). So as to use all features, a quarter of pixels are evenly selected from the supersampled image in order to use all pixels of the original data.</p

Research paper thumbnail of Unsupervised Learning in an Ensemble of Spiking Neural Networks Mediated by ITDP - Fig 12

Unsupervised learning in an ensemble of spiking neural networks mediated by ITDP Article (Publish... more Unsupervised learning in an ensemble of spiking neural networks mediated by ITDP Article (Published Version) http://sro.sussex.ac.uk Shim, Yoonsik, Philippides, Andy, Staras, Kevin and Husbands, Phil (2016) Unsupervised learning in an ensemble of spiking neural networks mediated by ITDP. PLoS Computational Biology, 12 (10). e1005137.

Research paper thumbnail of Spike trains from the SEM ensemble network with <i>N</i><sub><i>E</i></sub> = 5 and random feature selection

<p>(Left) Plot shows the input neuron spikes from eight image presentations from different ... more <p>(Left) Plot shows the input neuron spikes from eight image presentations from different classes (digits) which are depicted in different colors (black: 0, red: 1, green: 2, blue: 4). (Right) Two graphs show the output spikes of ensemble, gating, and final WTA neurons before and after learning. The colors of the spikes represent which class is being presented as input. After learning the network outputs produce consistent firing patterns, each output spiking exclusively for a single class.</p

Research paper thumbnail of Examples of random Gaussian mean placements for different <i>N</i><sub><i>E</i></sub> from the manually designed initial points (black points)

<p>The red pixels represent the outer border of the active region of the image, and the yel... more <p>The red pixels represent the outer border of the active region of the image, and the yellow pixels represent a forbidden region which is 3 pixels thick. The jittered mean points were restricted to be placed inside the inner region (including the green pixels) which is surrounded by the inner border (green).</p

Research paper thumbnail of Training performances of the expanded STDP/ITDP networks (using random feature selection on the MNIST handwritten digits classification task as in earlier experiments)

<p>Each color represents, red: ITDP final WTA, green: STDP final WTA, blue: gating WTA, gre... more <p>Each color represents, red: ITDP final WTA, green: STDP final WTA, blue: gating WTA, grey/black: ensemble WTAs and their average. (A, B) An example of time courses of performances and the final performances from 50 repeated trials using unsupervised gating WTA. The individual trials were sorted by gating WTA performances in ascending order. (C, D) Simulations using the automatic selection of gating WTA. The vertical lines with arrowheads in C indicate where the switching of gating WTA occurs (see text for further details).</p

Research paper thumbnail of A voter and the voter ensemble network (<i>N</i><sub><i>C</i></sub> = 4)

<p>(Left) A voter and the predefined firing probabilities of each voter neuron for a set of... more <p>(Left) A voter and the predefined firing probabilities of each voter neuron for a set of virtual input samples <i>X</i> = {<i>x</i><sub>1</sub>, <i>x</i><sub>2</sub>, …, <i>x</i><sub><i>M</i></sub>}. (Right) The voter ensemble network. The weight represents the weight of connection from the <i>i</i>th neuron of the <i>j</i>th voter to the <i>k</i>th neuron of the final voter.</p

Research paper thumbnail of 순응적 날개를 가진 날갯짓 로봇 시뮬레이션을 위한 뉴로컨트롤의 진화연산 최적화

The Korean Society Of Computer And Information, Dec 1, 2019

Research paper thumbnail of 경로 제어가 가능한 가상생명체를 위한 2단계 진화 알고리즘

Research paper thumbnail of Learning Media on Mathematical Education based on Augmented Reality

KSII Transactions on Internet and Information Systems, 2021

Modern technology offers many ways to enhance teaching and learning that in turn promote the deve... more Modern technology offers many ways to enhance teaching and learning that in turn promote the development of tools for educational activities both inside and outside the classroom. Many educational programs using the augmented reality (AR) technology are being widely used to provide supplementary learning materials for students. This paper describes the potential and challenges of using GeoGebra AR in mathematical studies, whereby students can view 3D geometric objects for a better understanding of their structure, and verifies the feasibility of its use based on experimental results. The GeoGebra software can be used to draw geometric objects, and 3D geometric objects can be viewed using AR software or AR applications on mobile phones or computer tablets. These could provide some of the required materials for mathematical education at high schools or universities. The use of the GeoGebra application for education in Laos will be particularly discussed in this paper.

Research paper thumbnail of Evolutionary Robotics and Neuroscience

The Horizons of Evolutionary Robotics, 2014

Research paper thumbnail of Generating flying creatures using body-brain co-evolution

This paper describes a system that produces double-winged flying creatures using body-brain co-ev... more This paper describes a system that produces double-winged flying creatures using body-brain co-evolution without need of complex flapping flight aerodynamics. While artificial life techniques have been used to create a variety of virtual creatures, little work has explored flapping-winged creatures for the difficulty of genetic encoding problem of wings with limited geometric primitives as well as flapping-wing aerodynamics. Despite of the simplicity of system, our result shows aesthetical looking and organic flapping flight locomotions. The restricted list structure is used in genotype encoding for morphological symmetry of creatures and is more easily handled than other data structures. The creatures evolved by this system have two symmetric flapping wings consisting of continuous triangular patches and show various looking and locomotion such as wings of birds, butterflies and bats or even imaginary wings of a dragon and pterosaurs.

Research paper thumbnail of Evolutionary Optimization of Neurocontroller for Physically Simulated Compliant-Wing Ornithopter

Journal of the Korea Society of Computer and Information, 2019

Research paper thumbnail of Recent advances in evolutionary and bio-inspired adaptive robotics: Exploiting embodied dynamics

Applied Intelligence, 2021

This paper explores current developments in evolutionary and bio-inspired approaches to autonomou... more This paper explores current developments in evolutionary and bio-inspired approaches to autonomous robotics, concentrating on research from our group at the University of Sussex. These developments are discussed in the context of advances in the wider fields of adaptive and evolutionary approaches to AI and robotics, focusing on the exploitation of embodied dynamics to create behaviour. Four case studies highlight various aspects of such exploitation. The first exploits the dynamical properties of a physical electronic substrate, demonstrating for the first time how component-level analog electronic circuits can be evolved directly in hardware to act as robot controllers. The second develops novel, effective and highly parsimonious navigation methods inspired by the way insects exploit the embodied dynamics of innate behaviours. Combining biological experiments with robotic modeling, it is shown how rapid route learning can be achieved with the aid of navigation-specific visual info...

Research paper thumbnail of Evolving Flying Creatures with Path Following Behaviors

Artificial Life IX, 2004

We present a system which evolves physically simulated 3D flying creatures and their maneuvers. T... more We present a system which evolves physically simulated 3D flying creatures and their maneuvers. The creature is modelled as a number of articulated cylinders connected by triangular patagia in between. A creature's wing structure and its low-level controllers for straight flight are generated by an evolutionary algorithm. Then a feed-forward neural network is attached to the low-level controllers, and the connection weights of the network for a given trajectory are found by a genetic algonthm. We show that a control system sufficiently effective to allow aerial creatures to follow a complicated path can be achieved by two-step evolution process.

Research paper thumbnail of Embodied neuromechanical chaos through homeostatic regulation

Chaos: An Interdisciplinary Journal of Nonlinear Science, 2019

In this paper, we present detailed analyses of the dynamics of a number of embodied neuromechanic... more In this paper, we present detailed analyses of the dynamics of a number of embodied neuromechanical systems of a class that has been shown to efficiently exploit chaos in the development and learning of motor behaviors for bodies of arbitrary morphology. This class of systems has been successfully used in robotics, as well as to model biological systems. At the heart of these systems are neural central pattern generating (CPG) units connected to actuators which return proprioceptive information via an adaptive homeostatic mechanism. Detailed dynamical analyses of example systems, using high resolution largest Lyapunov exponent maps, demonstrate the existence of chaotic regimes within a particular region of parameter space, as well as the striking similarity of the maps for systems of varying size. Thanks to the homeostatic sensory mechanisms, any single CPG “views” the whole of the rest of the system as if it was another CPG in a two coupled system, allowing a scale invariant concep...