Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146099
Ting Wang, Wing W. Y. Ng, Wendi Li, S. Kwong, Jingde Li
Deep Neural Networks (DNNs) demonstrate great performances in pattern classification problems. There are several available activation functions for DNNs while the Sigmoid and the Tanh functions are most widely used choices. In this work, we propose the Broad Autoencoder Features (BAF) to better utilize advantages of different activation functions. The BAF consists of four parallel connected Stacked AutoEncoders (SAEs) with different activation functions: the Sigmoid, the Tanh, the ReLu, and the Softplus. With this broad setting, the final learned features merge learn features using diversified nonlinear mappings from the original input features and such that more information is mined from the original input features. Experimental results show that the BAF yields better learned features in comparison with merging four SAEs using the same activation functions.
{"title":"Broad Autoencoder Features Learning for Pattern Classification Problems","authors":"Ting Wang, Wing W. Y. Ng, Wendi Li, S. Kwong, Jingde Li","doi":"10.1109/ICCICC46617.2019.9146099","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146099","url":null,"abstract":"Deep Neural Networks (DNNs) demonstrate great performances in pattern classification problems. There are several available activation functions for DNNs while the Sigmoid and the Tanh functions are most widely used choices. In this work, we propose the Broad Autoencoder Features (BAF) to better utilize advantages of different activation functions. The BAF consists of four parallel connected Stacked AutoEncoders (SAEs) with different activation functions: the Sigmoid, the Tanh, the ReLu, and the Softplus. With this broad setting, the final learned features merge learn features using diversified nonlinear mappings from the original input features and such that more information is mined from the original input features. Experimental results show that the BAF yields better learned features in comparison with merging four SAEs using the same activation functions.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121927327","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146062
K. Brezinski, K. Ferens
This paper presents a population based simulated annealing algorithm to improve modelling of cognitive processes. Particle Swarm Optimization (PSO) is embedded within the basic Simulated Annealing (SA) algorithm to allow for multiple concurrent candidate solutions through the use of a population-driven social coefficient updating the other population members. A modified ramping strategy which balances inertial, personal and swarm coefficients is introduced. The hybrid PSO/SA algorithm was tested on the travelling salesperson problem (TSP), and was shown to outperform the individual algorithms by improving their limitations in exploration and exploitation.
{"title":"Cognitive Hybrid PSO/SA Combinatorial Optimization","authors":"K. Brezinski, K. Ferens","doi":"10.1109/ICCICC46617.2019.9146062","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146062","url":null,"abstract":"This paper presents a population based simulated annealing algorithm to improve modelling of cognitive processes. Particle Swarm Optimization (PSO) is embedded within the basic Simulated Annealing (SA) algorithm to allow for multiple concurrent candidate solutions through the use of a population-driven social coefficient updating the other population members. A modified ramping strategy which balances inertial, personal and swarm coefficients is introduced. The hybrid PSO/SA algorithm was tested on the travelling salesperson problem (TSP), and was shown to outperform the individual algorithms by improving their limitations in exploration and exploitation.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121443091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146070
Shashanka Marigi Rajanarayana, Sumeet S. Kumar, A. Zjajo, R. V. Leuken
Advanced driving assistance systems (ADAS) prepave regulators, consumers and corporations for the medium-term reality of autonomous driving with adaptive cruise control, collision avoidance and lane departure warning system. Various sensors like camera, RADAR and LIDAR, integrated into the vehicle assist driving. In addition, deep learning approaches are utilized in a wide range of applications ranging from object detection and scene segmentation to engine fault diagnosis and emission management to detect vehicle network intrusion. In this paper, we scope out the state of the art sensors subsystems in terms of its functionality, characteristics, specifications and communication protocol, and we describe cognitive deep learning based algorithms required for environment perception through these sensors. Subsequently, we analyze the cognitive algorithm by profiling the standard deep learning models, explore different compute platforms and possible algorithm and hardware optimization scenarios.
{"title":"Towards Computationally-Efficient Cognitive Sensor Systems for Autonomous Vehicles","authors":"Shashanka Marigi Rajanarayana, Sumeet S. Kumar, A. Zjajo, R. V. Leuken","doi":"10.1109/ICCICC46617.2019.9146070","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146070","url":null,"abstract":"Advanced driving assistance systems (ADAS) prepave regulators, consumers and corporations for the medium-term reality of autonomous driving with adaptive cruise control, collision avoidance and lane departure warning system. Various sensors like camera, RADAR and LIDAR, integrated into the vehicle assist driving. In addition, deep learning approaches are utilized in a wide range of applications ranging from object detection and scene segmentation to engine fault diagnosis and emission management to detect vehicle network intrusion. In this paper, we scope out the state of the art sensors subsystems in terms of its functionality, characteristics, specifications and communication protocol, and we describe cognitive deep learning based algorithms required for environment perception through these sensors. Subsequently, we analyze the cognitive algorithm by profiling the standard deep learning models, explore different compute platforms and possible algorithm and hardware optimization scenarios.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132958640","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146060
Zartasha Mustansar, Maria Rathore, A. Shaukat, Faizan Nadeem, Nabisha Farooq, Salma Sherbaz
Developing geometries of the real objects using computer aided engineering methods has been a common practice now. However, due to the evolution of advancement and launch of digital age, there is a recent interest to develop refined, smooth and entirely significant geometrical details to account for accuracy in predictions. Whether it is a scientific computation or reverse engineering, simulation or geometrical reconstruction; data sets with delicate geometric details are created quite often for various purposes. The usefulness of such geometric details rests on the ability to process them efficiently i.e. from digital models to numerical models and eventually for high-end visualization data analysis. In the field of biomedical engineering, geometry plays a very important role in model prediction. This study therefore considers the significance of geometry in the human head model to calculate critical pressure in the brain named “Intra-Cranial Pressure”. Elevated intracranial pressure (ICP) is one of the common consequences of traumatic conditions and has a profound influence on outcome. There are well established methods for the measurement, continuous monitoring and treatment of raised ICP. However, there is a need to build computer models for the same for validation and prediction. We made use of a tumour brain, in this study to see how geometry varies the values acquired for ICP in the brain. One of the major benefits of this study will be non-invasive computation of pressure inside the brain in a safe frequency range. It is well established that the relation between volume and pressure is non-linear. Additionally, skull is usually, considered as an enclosed and in-elastic container like a sac. The positioning of layers within this sac generates a constant pressure which is normal according to the body homeostasis. An increase in the volume of any of the intra cranial contents (Sac contents) is naturally offset by a decrease in pressure in one or the other content in it. However, when the size of the tumor (which is not an intracranial content) increases, the compensatory mechanisms gets exhausted and further increase in the brain sac in terms of volume results in an extremely elevated ICP. This mechanism is replicated in this research by using two approaches based on geometry: (i) Simple Geometry using Image based Finite Element modeling (ii) A regular engineering geometry using CAD modeling in Abaqus CAE. Reportedly the normal range of ICP lies between 3.75~15mmHg in humans. We have generated two head models with these approaches using the same boundary conditions and loading parameters.
{"title":"Geometrical Vitality of Human Head model to Calculate Intra Cranial Pressure for Procognitive Computing","authors":"Zartasha Mustansar, Maria Rathore, A. Shaukat, Faizan Nadeem, Nabisha Farooq, Salma Sherbaz","doi":"10.1109/ICCICC46617.2019.9146060","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146060","url":null,"abstract":"Developing geometries of the real objects using computer aided engineering methods has been a common practice now. However, due to the evolution of advancement and launch of digital age, there is a recent interest to develop refined, smooth and entirely significant geometrical details to account for accuracy in predictions. Whether it is a scientific computation or reverse engineering, simulation or geometrical reconstruction; data sets with delicate geometric details are created quite often for various purposes. The usefulness of such geometric details rests on the ability to process them efficiently i.e. from digital models to numerical models and eventually for high-end visualization data analysis. In the field of biomedical engineering, geometry plays a very important role in model prediction. This study therefore considers the significance of geometry in the human head model to calculate critical pressure in the brain named “Intra-Cranial Pressure”. Elevated intracranial pressure (ICP) is one of the common consequences of traumatic conditions and has a profound influence on outcome. There are well established methods for the measurement, continuous monitoring and treatment of raised ICP. However, there is a need to build computer models for the same for validation and prediction. We made use of a tumour brain, in this study to see how geometry varies the values acquired for ICP in the brain. One of the major benefits of this study will be non-invasive computation of pressure inside the brain in a safe frequency range. It is well established that the relation between volume and pressure is non-linear. Additionally, skull is usually, considered as an enclosed and in-elastic container like a sac. The positioning of layers within this sac generates a constant pressure which is normal according to the body homeostasis. An increase in the volume of any of the intra cranial contents (Sac contents) is naturally offset by a decrease in pressure in one or the other content in it. However, when the size of the tumor (which is not an intracranial content) increases, the compensatory mechanisms gets exhausted and further increase in the brain sac in terms of volume results in an extremely elevated ICP. This mechanism is replicated in this research by using two approaches based on geometry: (i) Simple Geometry using Image based Finite Element modeling (ii) A regular engineering geometry using CAD modeling in Abaqus CAE. Reportedly the normal range of ICP lies between 3.75~15mmHg in humans. We have generated two head models with these approaches using the same boundary conditions and loading parameters.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133385231","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146104
Oussama Sabri, A. Muzy
Learning dynamics at cognitive process level is difficult to study and emulate because of the complexity of intricate psychological and neuronal mechanisms and dynamics. When considering the parallel processing of a task, the difficulty relies on the execution concurrency making the process contributions indistinguishable. We present here a metric for rewarding increasingly the right parallel cognitive processes with respect to the wrong ones through learning steps. The metric, based on the symmetric difference between task parallel processes, proves to correctly achieve collective and individual credit assignment of the processes.
{"title":"Learning Dynamics of Cognitive Parallel Processing Based on a Collective Evaluation","authors":"Oussama Sabri, A. Muzy","doi":"10.1109/ICCICC46617.2019.9146104","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146104","url":null,"abstract":"Learning dynamics at cognitive process level is difficult to study and emulate because of the complexity of intricate psychological and neuronal mechanisms and dynamics. When considering the parallel processing of a task, the difficulty relies on the execution concurrency making the process contributions indistinguishable. We present here a metric for rewarding increasingly the right parallel cognitive processes with respect to the wrong ones through learning steps. The metric, based on the symmetric difference between task parallel processes, proves to correctly achieve collective and individual credit assignment of the processes.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"103 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124571662","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146076
U. Segerstrale
It has been recently realized that the current higher education does not adequately prepare students for the jobs of the future. Rather than narrow specialization, what will be needed are skills and attitudes that vouch for a smooth transition from school to work place and the ability of a young person to further develop but also adapt to the demands of the job. Many universities are now looking to develop basic skill sets which emphasize such things as communication, team work, creativity, and crossdisciplinary competence. This is specially the case for technically oriented schools, whose students will enter a world which favors collaboration-driven innovation, currently regarded as the best way to rapid development. While the current university curriculum still invites relatively passive learning, new initiatives have been taken for such things as creativity workshops, or faculty development seminars for re-imagining education. One recent experiment at my own university was an Artificial Intelligence Collaboration Day with “flash presentations” by students and faculty as well as longer presentations and group discussions. The idea was for people from widely different fields to identify common interests for potential collaboration, and this worked well because of the friendly atmosphere. The most tangible current experiments are specially built “innovation centers”, such as the new Kaplan Institute for Innovation at Illinois Tech - a building which is specially designed for innovation through collaboration. Flexible architecture and new interior design can quickly adapt to the needs of different projects and audiences. The biggest challenge, however, and a key concern for the education of the future, is creating a learning climate where quiet individual students can develop into happily communicating, competent and confident human beings. I will mention some of my own experiments in this respect within the American university system and finally take a look at a surprising but functioning alternative: the Finnish educational system and its underlying values.
{"title":"Education for Creativity, Skills, and Cross-disciplinary Collaboration","authors":"U. Segerstrale","doi":"10.1109/ICCICC46617.2019.9146076","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146076","url":null,"abstract":"It has been recently realized that the current higher education does not adequately prepare students for the jobs of the future. Rather than narrow specialization, what will be needed are skills and attitudes that vouch for a smooth transition from school to work place and the ability of a young person to further develop but also adapt to the demands of the job. Many universities are now looking to develop basic skill sets which emphasize such things as communication, team work, creativity, and crossdisciplinary competence. This is specially the case for technically oriented schools, whose students will enter a world which favors collaboration-driven innovation, currently regarded as the best way to rapid development. While the current university curriculum still invites relatively passive learning, new initiatives have been taken for such things as creativity workshops, or faculty development seminars for re-imagining education. One recent experiment at my own university was an Artificial Intelligence Collaboration Day with “flash presentations” by students and faculty as well as longer presentations and group discussions. The idea was for people from widely different fields to identify common interests for potential collaboration, and this worked well because of the friendly atmosphere. The most tangible current experiments are specially built “innovation centers”, such as the new Kaplan Institute for Innovation at Illinois Tech - a building which is specially designed for innovation through collaboration. Flexible architecture and new interior design can quickly adapt to the needs of different projects and audiences. The biggest challenge, however, and a key concern for the education of the future, is creating a learning climate where quiet individual students can develop into happily communicating, competent and confident human beings. I will mention some of my own experiments in this respect within the American university system and finally take a look at a surprising but functioning alternative: the Finnish educational system and its underlying values.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"2016 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114655242","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146028
Jun Peng, Shaoning Pang, Du Zhang, Shangzhu Jin, Lixiao Feng, Zuojin Li
For a security system built on symmetric-key cryptography algorithms, the substitution box (S-box) plays a crucial role to resist cryptanalysis (decoding). In this paper, we incorporate quantum chaos and PWLCM chaotic mapping into a new method of S-box design. Over the obtained 500 key-dependent S-boxes, we test the S-box cryptographical properties on bijection, nonlinearity, SAC, BIC, differential approximation probability, and sensitivity to the key, respectively. The results show that the cryptographic characteristics of proposed S-Box has met our design objectives and can be applied to data encryption, user authentication and system access control.
{"title":"S-boxes Construction Based on Quantum Chaos and PWLCM Chaotic Mapping","authors":"Jun Peng, Shaoning Pang, Du Zhang, Shangzhu Jin, Lixiao Feng, Zuojin Li","doi":"10.1109/ICCICC46617.2019.9146028","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146028","url":null,"abstract":"For a security system built on symmetric-key cryptography algorithms, the substitution box (S-box) plays a crucial role to resist cryptanalysis (decoding). In this paper, we incorporate quantum chaos and PWLCM chaotic mapping into a new method of S-box design. Over the obtained 500 key-dependent S-boxes, we test the S-box cryptographical properties on bijection, nonlinearity, SAC, BIC, differential approximation probability, and sensitivity to the key, respectively. The results show that the cryptographic characteristics of proposed S-Box has met our design objectives and can be applied to data encryption, user authentication and system access control.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117042251","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146079
J. C. V. Tieck, Pascal Becker, Jacques Kaiser, Igor Peric, Mahmoud Akl, Daniel Reichard, A. Rönnau, R. Dillmann
The main purpose of the human arm is to reach a target and perform a manipulation task. Human babies learn to move their arms by imitating and doing motor babbling through trial and error. This learning is believed to result from changes in synaptic efficacy triggered by complex mechanisms involving neuro-modulators in which dopamine plays a key role. After learning, humans are able to reuse and adapt the motions without performing complex calculations. In contrast, classical robotics achieve target reaching by mathematically computing each time the inverse kinematics (IK) of the joint angles leading to a particular target, then validating the configuration and generating a trajectory. This process is computational intensive and becomes more complex with the amount of degrees of freedom (DoF). In this work, we propose a spiking neural network architecture to learn target reaching motions with a robotic arm using reinforcement learning (RL), which is closely related to the way babies learn. To make our approach scalable, we sub-divide the kinematics structure of the robot and create one sub-network per joint. We generate training data offline by generating random reaching motions with an IK calculation outside of the network. After learning, the IK is no longer required, and the model is implicitly learned in the weights of the network. Mimicking the learning mechanisms of the brain, we use the spike time dependent plasticity (STDP) learning rule modulated by dopamine, representing a reward. The approach is evaluated with a simulated Universal Robot UR5 with six DoF. The network successfully learns to reach multiple targets and by changing the reward function on-the-fly it is able to learn different control functions. With a standard computer our network was able to control a robotic kinematics chain up to 13 DoF in real time. A key aspect of our approach is that in contrast to deep RL our SNN does not need much data to learn new behaviors. We believe that model free motion controllers inspired on the human brain mechanisms can improve the way robots are programmed by making the process more adaptive and flexible.
人的手臂的主要目的是达到目标并执行操作任务。人类婴儿通过模仿和反复试验来学习移动手臂。这种学习被认为是由涉及神经调节剂的复杂机制引发的突触效能变化引起的，其中多巴胺起着关键作用。在学习之后，人类能够重复使用和适应这些动作，而无需进行复杂的计算。而经典机器人通过数学计算每次指向特定目标的关节角的逆运动学(IK)，然后验证构型并生成轨迹来实现目标到达。这个过程是计算密集型的，并且随着自由度(DoF)的增加变得更加复杂。在这项工作中，我们提出了一个尖峰神经网络架构，使用强化学习(RL)来学习机械臂的目标到达运动，这与婴儿的学习方式密切相关。为了使我们的方法具有可扩展性，我们对机器人的运动学结构进行细分，并为每个关节创建一个子网络。我们通过在网络外使用IK计算生成随机到达运动来离线生成训练数据。学习后，不再需要IK，模型在网络的权值中隐式学习。模仿大脑的学习机制，我们使用由多巴胺调节的spike time dependent plasticity (STDP)学习规则，代表一种奖励。用仿真的六自由度通用机器人UR5对该方法进行了验证。该网络成功地学会了到达多个目标，并且通过动态改变奖励函数，它能够学习不同的控制函数。在一台标准的计算机上，我们的网络能够实时控制机器人运动链达到13自由度。我们方法的一个关键方面是，与深度强化学习相比，我们的SNN不需要太多数据来学习新的行为。我们相信，受人脑机制启发的无模型运动控制器可以通过使过程更具适应性和灵活性来改进机器人的编程方式。
{"title":"Learning target reaching motions with a robotic arm using brain-inspired dopamine modulated STDP","authors":"J. C. V. Tieck, Pascal Becker, Jacques Kaiser, Igor Peric, Mahmoud Akl, Daniel Reichard, A. Rönnau, R. Dillmann","doi":"10.1109/ICCICC46617.2019.9146079","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146079","url":null,"abstract":"The main purpose of the human arm is to reach a target and perform a manipulation task. Human babies learn to move their arms by imitating and doing motor babbling through trial and error. This learning is believed to result from changes in synaptic efficacy triggered by complex mechanisms involving neuro-modulators in which dopamine plays a key role. After learning, humans are able to reuse and adapt the motions without performing complex calculations. In contrast, classical robotics achieve target reaching by mathematically computing each time the inverse kinematics (IK) of the joint angles leading to a particular target, then validating the configuration and generating a trajectory. This process is computational intensive and becomes more complex with the amount of degrees of freedom (DoF). In this work, we propose a spiking neural network architecture to learn target reaching motions with a robotic arm using reinforcement learning (RL), which is closely related to the way babies learn. To make our approach scalable, we sub-divide the kinematics structure of the robot and create one sub-network per joint. We generate training data offline by generating random reaching motions with an IK calculation outside of the network. After learning, the IK is no longer required, and the model is implicitly learned in the weights of the network. Mimicking the learning mechanisms of the brain, we use the spike time dependent plasticity (STDP) learning rule modulated by dopamine, representing a reward. The approach is evaluated with a simulated Universal Robot UR5 with six DoF. The network successfully learns to reach multiple targets and by changing the reward function on-the-fly it is able to learn different control functions. With a standard computer our network was able to control a robotic kinematics chain up to 13 DoF in real time. A key aspect of our approach is that in contrast to deep RL our SNN does not need much data to learn new behaviors. We believe that model free motion controllers inspired on the human brain mechanisms can improve the way robots are programmed by making the process more adaptive and flexible.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128747587","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/ICCICC46617.2019.9146037
Janani Ramanathan
This paper explores the potential of the rapidly evolving fields of Natural Language Processing and Affective Computing and proposes future applications that combine the power of both fields to assist individuals in their personal and collective accomplishment. It studies the latest developments in the field of Emotion Detection and Recognition from facial expression, voice and text and discusses the shortcomings in current analysis systems. Human subjectivity is key to every choice, decision and act of individuals, and a comprehensive knowledge of human psychology is essential for effective analysis. As Emotional AI transcends the physical parameters and moves closer to understanding the emotional and mental human being in future, and Deep Learning enables greater comprehension of unstructured textual and audio-visual data, Cognitive Computing can employ big data processing to assist humans in acquiring scholarship, anticipating social trends and even understanding life. The paper concludes with a proposal for a revolutionary field of Artificial Life Intelligence that can promote universal human welfare.
{"title":"Artificial Life Intelligence for Individual and Societal Accomplishment","authors":"Janani Ramanathan","doi":"10.1109/ICCICC46617.2019.9146037","DOIUrl":"https://doi.org/10.1109/ICCICC46617.2019.9146037","url":null,"abstract":"This paper explores the potential of the rapidly evolving fields of Natural Language Processing and Affective Computing and proposes future applications that combine the power of both fields to assist individuals in their personal and collective accomplishment. It studies the latest developments in the field of Emotion Detection and Recognition from facial expression, voice and text and discusses the shortcomings in current analysis systems. Human subjectivity is key to every choice, decision and act of individuals, and a comprehensive knowledge of human psychology is essential for effective analysis. As Emotional AI transcends the physical parameters and moves closer to understanding the emotional and mental human being in future, and Deep Learning enables greater comprehension of unstructured textual and audio-visual data, Cognitive Computing can employ big data processing to assist humans in acquiring scholarship, anticipating social trends and even understanding life. The paper concludes with a proposal for a revolutionary field of Artificial Life Intelligence that can promote universal human welfare.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132905210","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/iccicc46617.2019.9146082
E. Tunstel
Today's discourse among technical professionals and technology enthusiasts alike is teeming with subject matter focused on innovations resulting from the research and practice of systems science and engineering, human-machine systems, and cybernetics. Whether it is complex systems enabled by cybernetics, intelligence for robotic and vehicular autonomy, new capabilities enabled by advances in machine learning, augmented humans, human-machine fusion, or other forms of human-machine symbiosis, the dialog is vibrant in technical and non-technical sectors of society. The convergence of these focal areas is prevalent at the current cutting edge of technology, but with a more pronounced emphasis on human factors and human relationships to technologies comprising complex systems and toward enabling appropriate human-centric solutions. With cybernetics as a science of, and transdisciplinary approach to studying, control and communications in machines and living things, its elements can be combined to enable complex and increasingly intelligent systems that interact with humans in a symbiotic or collaborative fashion. This talk focuses on such systems in the form of intelligent or otherwise cognitive robots. In that context, it highlights applications involving ideas from cybernetics and human-robot interaction research, considerations for next-level robotic intelligence needed to enable smart human-collaborative robots, and opportunities for leveraging transdisciplinary ideas that would enhance such robotic systems.
{"title":"Transdisciplinary Convergence of Human-Centric Robotic Systems and Cybernetics","authors":"E. Tunstel","doi":"10.1109/iccicc46617.2019.9146082","DOIUrl":"https://doi.org/10.1109/iccicc46617.2019.9146082","url":null,"abstract":"Today's discourse among technical professionals and technology enthusiasts alike is teeming with subject matter focused on innovations resulting from the research and practice of systems science and engineering, human-machine systems, and cybernetics. Whether it is complex systems enabled by cybernetics, intelligence for robotic and vehicular autonomy, new capabilities enabled by advances in machine learning, augmented humans, human-machine fusion, or other forms of human-machine symbiosis, the dialog is vibrant in technical and non-technical sectors of society. The convergence of these focal areas is prevalent at the current cutting edge of technology, but with a more pronounced emphasis on human factors and human relationships to technologies comprising complex systems and toward enabling appropriate human-centric solutions. With cybernetics as a science of, and transdisciplinary approach to studying, control and communications in machines and living things, its elements can be combined to enable complex and increasingly intelligent systems that interact with humans in a symbiotic or collaborative fashion. This talk focuses on such systems in the form of intelligent or otherwise cognitive robots. In that context, it highlights applications involving ideas from cybernetics and human-robot interaction research, considerations for next-level robotic intelligence needed to enable smart human-collaborative robots, and opportunities for leveraging transdisciplinary ideas that would enhance such robotic systems.","PeriodicalId":294902,"journal":{"name":"2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134639303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：