Publications
Search
21.
Ustun, Volkan; Rosenbloom, Paul S; Sajjadi, Seyed; Nuttall, Jeremy
Controlling Synthetic Characters in Simulations: A Case for Cognitive Architectures and Sigma Proceedings Article
In: Proceedings of I/ITSEC 2018, National Training and Simulation Association, Orlando, FL, 2018.
@inproceedings{ustun_controlling_2018,
title = {Controlling Synthetic Characters in Simulations: A Case for Cognitive Architectures and Sigma},
author = {Volkan Ustun and Paul S Rosenbloom and Seyed Sajjadi and Jeremy Nuttall},
url = {http://bcf.usc.edu/ rosenblo/Pubs/Ustun_IITSEC2018_D.pdf},
year = {2018},
date = {2018-11-01},
booktitle = {Proceedings of I/ITSEC 2018},
publisher = {National Training and Simulation Association},
address = {Orlando, FL},
abstract = {Simulations, along with other similar applications like virtual worlds and video games, require computational models of intelligence that generate realistic and credible behavior for the participating synthetic characters. Cognitive architectures, which are models of the fixed structure underlying intelligent behavior in both natural and artificial systems, provide a conceptually valid common basis, as evidenced by the current efforts towards a standard model of the mind, to generate human-like intelligent behavior for these synthetic characters. Developments in the field of artificial intelligence, mainly in probabilistic graphical models and neural networks, open up new opportunities for cognitive architectures to make the synthetic characters more autonomous and to enrich their behavior. Sigma (Σ) is a cognitive architecture and system that strives to combine what has been learned from four decades of independent work on symbolic cognitive architectures, probabilistic graphical models, and more recently neural models, under its graphical architecture hypothesis. Sigma leverages an extended form of factor graphs towards a uniform grand unification of not only traditional cognitive capabilities but also key non-cognitive aspects, creating unique opportunities for the construction of new kinds of cognitive models that possess a Theory-of-Mind and that are perceptual, autonomous, interactive, affective, and adaptive. In this paper, we will introduce Sigma along with its diverse capabilities and then use three distinct proof-of-concept Sigma models to highlight combinations of these capabilities: (1) Distributional reinforcement learning models in a simple OpenAI Gym problem; (2) A pair of adaptive and interactive agent models that demonstrate rule-based, probabilistic, and social reasoning in a physical security scenario instantiated within the SmartBody character animation platform; and (3) A knowledge-free exploration model in which an agent leverages only architectural appraisal variables, namely attention and curiosity, to locate an item while building up a map in a Unity environment.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Simulations, along with other similar applications like virtual worlds and video games, require computational models of intelligence that generate realistic and credible behavior for the participating synthetic characters. Cognitive architectures, which are models of the fixed structure underlying intelligent behavior in both natural and artificial systems, provide a conceptually valid common basis, as evidenced by the current efforts towards a standard model of the mind, to generate human-like intelligent behavior for these synthetic characters. Developments in the field of artificial intelligence, mainly in probabilistic graphical models and neural networks, open up new opportunities for cognitive architectures to make the synthetic characters more autonomous and to enrich their behavior. Sigma (Σ) is a cognitive architecture and system that strives to combine what has been learned from four decades of independent work on symbolic cognitive architectures, probabilistic graphical models, and more recently neural models, under its graphical architecture hypothesis. Sigma leverages an extended form of factor graphs towards a uniform grand unification of not only traditional cognitive capabilities but also key non-cognitive aspects, creating unique opportunities for the construction of new kinds of cognitive models that possess a Theory-of-Mind and that are perceptual, autonomous, interactive, affective, and adaptive. In this paper, we will introduce Sigma along with its diverse capabilities and then use three distinct proof-of-concept Sigma models to highlight combinations of these capabilities: (1) Distributional reinforcement learning models in a simple OpenAI Gym problem; (2) A pair of adaptive and interactive agent models that demonstrate rule-based, probabilistic, and social reasoning in a physical security scenario instantiated within the SmartBody character animation platform; and (3) A knowledge-free exploration model in which an agent leverages only architectural appraisal variables, namely attention and curiosity, to locate an item while building up a map in a Unity environment.
22.
Rosenbloom, Paul S.; Demski, Abram; Ustun, Volkan
Toward a Neural-Symbolic Sigma: Introducing Neural Network Learning Proceedings Article
In: Proceedings of the 15th Annual Meeting of the International Conference on Cognitive Modelling, 2002–2017 EasyChair, Coventry, United Kingdom, 2017.
@inproceedings{rosenbloom_toward_2017,
title = {Toward a Neural-Symbolic Sigma: Introducing Neural Network Learning},
author = {Paul S. Rosenbloom and Abram Demski and Volkan Ustun},
url = {http://cs.usc.edu/ rosenblo/Pubs/ESNNL%20D.pdf},
year = {2017},
date = {2017-07-01},
booktitle = {Proceedings of the 15th Annual Meeting of the International Conference on Cognitive Modelling},
publisher = {2002–2017 EasyChair},
address = {Coventry, United Kingdom},
abstract = {Building on earlier work extending Sigma’s mixed (symbols + probabilities) graphical band to inference in feedforward neural networks, two forms of neural network learning – target propagation and backpropagation – are introduced, bringing Sigma closer to a full neural-symbolic architecture. Adapting Sigma’s reinforcement learning (RL) capability to use neural networks in policy learning then yields a hybrid form of neural RL with probabilistic action modeling.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Building on earlier work extending Sigma’s mixed (symbols + probabilities) graphical band to inference in feedforward neural networks, two forms of neural network learning – target propagation and backpropagation – are introduced, bringing Sigma closer to a full neural-symbolic architecture. Adapting Sigma’s reinforcement learning (RL) capability to use neural networks in policy learning then yields a hybrid form of neural RL with probabilistic action modeling.
23.
Joshi, Himanshu; Rosenbloom, Paul S.; Ustun, Volkan
Continuous phone recognition in the Sigma cognitive architecture Journal Article
In: Biologically Inspired Cognitive Architectures, vol. 18, pp. 23–32, 2016, ISSN: 2212683X.
@article{joshi_continuous_2016,
title = {Continuous phone recognition in the Sigma cognitive architecture},
author = {Himanshu Joshi and Paul S. Rosenbloom and Volkan Ustun},
url = {http://linkinghub.elsevier.com/retrieve/pii/S2212683X16300652},
doi = {10.1016/j.bica.2016.09.001},
issn = {2212683X},
year = {2016},
date = {2016-10-01},
journal = {Biologically Inspired Cognitive Architectures},
volume = {18},
pages = {23–32},
abstract = {Spoken language processing is an important capability of human intelligence that has hitherto been unexplored by cognitive architectures. This reflects on both the symbolic and sub-symbolic nature of the speech problem, and the capabilities provided by cognitive architectures to model the latter and its rich interplay with the former. Sigma has been designed to leverage the state-of-the-art hybrid (discrete + continuous) mixed (symbolic + probabilistic) capability of graphical models to provide in a uniform non-modular fashion effective forms of, and integration across, both cognitive and sub-cognitive behavior. In this article, previous work on speaker dependent isolated word recognition has been extended to demonstrate Sigma’s feasibility to process a stream of fluent audio and recognize phones, in an online and incremental manner with speaker independence. Phone recognition is an important step in integrating spoken language processing into Sigma. This work also extends the acoustic front-end used in the previous work in service of speaker independence. All of the knowledge used in phone recognition was added supraarchitecturally – i.e. on top of the architecture – without requiring the addition of new mechanisms to the architecture.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Spoken language processing is an important capability of human intelligence that has hitherto been unexplored by cognitive architectures. This reflects on both the symbolic and sub-symbolic nature of the speech problem, and the capabilities provided by cognitive architectures to model the latter and its rich interplay with the former. Sigma has been designed to leverage the state-of-the-art hybrid (discrete + continuous) mixed (symbolic + probabilistic) capability of graphical models to provide in a uniform non-modular fashion effective forms of, and integration across, both cognitive and sub-cognitive behavior. In this article, previous work on speaker dependent isolated word recognition has been extended to demonstrate Sigma’s feasibility to process a stream of fluent audio and recognize phones, in an online and incremental manner with speaker independence. Phone recognition is an important step in integrating spoken language processing into Sigma. This work also extends the acoustic front-end used in the previous work in service of speaker independence. All of the knowledge used in phone recognition was added supraarchitecturally – i.e. on top of the architecture – without requiring the addition of new mechanisms to the architecture.
24.
Rosenbloom, Paul S.; Demski, Abram; Ustun, Volkan
The Sigma Cognitive Architecture and System: Towards Functionally Elegant Grand Unification Journal Article
In: Journal of Artificial General Intelligence, 2016, ISSN: 1946-0163.
@article{rosenbloom_sigma_2016,
title = {The Sigma Cognitive Architecture and System: Towards Functionally Elegant Grand Unification},
author = {Paul S. Rosenbloom and Abram Demski and Volkan Ustun},
url = {http://www.degruyter.com/view/j/jagi.ahead-of-print/jagi-2016-0001/jagi-2016-0001.xml},
doi = {10.1515/jagi-2016-0001},
issn = {1946-0163},
year = {2016},
date = {2016-07-01},
journal = {Journal of Artificial General Intelligence},
abstract = {Sigma (Σ) is a cognitive architecture and system whose development is driven by a combination of four desiderata: grand unification, generic cognition, functional elegance, and sufficient efficiency. Work towards these desiderata is guided by the graphical architecture hypothesis, that key to progress on them is combining what has been learned from over three decades’ worth of separate work on cognitive architectures and graphical models. In this article, these four desiderata are motivated and explained, and then combined with the graphical architecture hypothesis to yield a rationale for the development of Sigma. The current state of the cognitive architecture is then introduced in detail, along with the graphical architecture that sits below it and implements it. Progress in extending Sigma beyond these architectures and towards a full cognitive system is then detailed in terms of both a systematic set of higher level cognitive idioms that have been developed and several virtual humans that are built from combinations of these idioms. Sigma as a whole is then analyzed in terms of how well the progress to date satisfies the desiderata. This article thus provides the first full motivation, presentation and analysis of Sigma, along with a diversity of more specific results that have been generated during its development.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Sigma (Σ) is a cognitive architecture and system whose development is driven by a combination of four desiderata: grand unification, generic cognition, functional elegance, and sufficient efficiency. Work towards these desiderata is guided by the graphical architecture hypothesis, that key to progress on them is combining what has been learned from over three decades’ worth of separate work on cognitive architectures and graphical models. In this article, these four desiderata are motivated and explained, and then combined with the graphical architecture hypothesis to yield a rationale for the development of Sigma. The current state of the cognitive architecture is then introduced in detail, along with the graphical architecture that sits below it and implements it. Progress in extending Sigma beyond these architectures and towards a full cognitive system is then detailed in terms of both a systematic set of higher level cognitive idioms that have been developed and several virtual humans that are built from combinations of these idioms. Sigma as a whole is then analyzed in terms of how well the progress to date satisfies the desiderata. This article thus provides the first full motivation, presentation and analysis of Sigma, along with a diversity of more specific results that have been generated during its development.
25.
Rosenbloom, Paul S.; Demski, Abram; Ustun, Volkan
Rethinking Sigma’s Graphical Architecture: An Extension to Neural Networks Proceedings Article
In: International Conference on Artificial General Intelligence, pp. 84–94, Springer, New York, NY, 2016, ISBN: 978-3-319-41649-6.
@inproceedings{rosenbloom_rethinking_2016,
title = {Rethinking Sigma’s Graphical Architecture: An Extension to Neural Networks},
author = {Paul S. Rosenbloom and Abram Demski and Volkan Ustun},
url = {http://link.springer.com/chapter/10.1007/978-3-319-41649-6_9},
doi = {10.1007/978-3-319-41649-6_9},
isbn = {978-3-319-41649-6},
year = {2016},
date = {2016-07-01},
booktitle = {International Conference on Artificial General Intelligence},
volume = {9782},
pages = {84–94},
publisher = {Springer},
address = {New York, NY},
abstract = {The status of Sigma’s grounding in graphical models is challenged by the ways in which their semantics has been violated while incorporating rule-based reasoning into them. This has led to a rethinking of what goes on in its graphical architecture, with results that include a straightforward extension to feedforward neural networks (although not yet with learning).},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The status of Sigma’s grounding in graphical models is challenged by the ways in which their semantics has been violated while incorporating rule-based reasoning into them. This has led to a rethinking of what goes on in its graphical architecture, with results that include a straightforward extension to feedforward neural networks (although not yet with learning).
26.
Ustun, Volkan; Rosenbloom, Paul
Towards Truly Autonomous Synthetic Characters with the Sigma Cognitive Architecture Book Section
In: Integrating Cognitive Architectures into Virtual Character Design, pp. 213 – 237, IGI Global, Hershey, PA, 2016, ISBN: 978-1-5225-0454-2.
@incollection{ustun_towards_2016,
title = {Towards Truly Autonomous Synthetic Characters with the Sigma Cognitive Architecture},
author = {Volkan Ustun and Paul Rosenbloom},
url = {http://services.igi-global.com/resolvedoi/resolve.aspx?doi=10.4018/978-1-5225-0454-2},
isbn = {978-1-5225-0454-2},
year = {2016},
date = {2016-06-01},
booktitle = {Integrating Cognitive Architectures into Virtual Character Design},
pages = {213 – 237},
publisher = {IGI Global},
address = {Hershey, PA},
abstract = {Realism is required not only for how synthetic characters look but also for how they behave. Many applications, such as simulations, virtual worlds, and video games, require computational models of intelligence that generate realistic and credible behavior for the participating synthetic characters. Sigma (Σ) is being built as a computational model of general intelligence with a long-term goal of understanding and replicating the architecture of the mind; i.e., the fixed structure underlying intelligent behavior. Sigma leverages probabilistic graphical models towards a uniform grand unification of not only traditional cognitive capabilities but also key non-cognitive aspects, creating unique opportunities for the construction of new kinds of non-modular behavioral models. These ambitions strive for the complete control of synthetic characters that behave as humanly as possible. In this paper, Sigma is introduced along with two disparate proof-of-concept virtual humans – one conversational and the other a pair of ambulatory agents – that demonstrate its diverse capabilities.},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Realism is required not only for how synthetic characters look but also for how they behave. Many applications, such as simulations, virtual worlds, and video games, require computational models of intelligence that generate realistic and credible behavior for the participating synthetic characters. Sigma (Σ) is being built as a computational model of general intelligence with a long-term goal of understanding and replicating the architecture of the mind; i.e., the fixed structure underlying intelligent behavior. Sigma leverages probabilistic graphical models towards a uniform grand unification of not only traditional cognitive capabilities but also key non-cognitive aspects, creating unique opportunities for the construction of new kinds of non-modular behavioral models. These ambitions strive for the complete control of synthetic characters that behave as humanly as possible. In this paper, Sigma is introduced along with two disparate proof-of-concept virtual humans – one conversational and the other a pair of ambulatory agents – that demonstrate its diverse capabilities.
27.
Ustun, Volkan; Rosenbloom, Paul S.; Kim, Julia; Li, Lingshan
BUILDING HIGH FIDELITY HUMAN BEHAVIOR MODELS IN THE SIGMA COGNITIVE ARCHITECTURE Proceedings Article
In: Proceedings of the 2015 Winter Simulation Conference, pp. 3124–3125, IEEE, Huntington Beach, CA, 2015, ISBN: 978-1-4673-9741-4.
@inproceedings{ustun_building_2015,
title = {BUILDING HIGH FIDELITY HUMAN BEHAVIOR MODELS IN THE SIGMA COGNITIVE ARCHITECTURE},
author = {Volkan Ustun and Paul S. Rosenbloom and Julia Kim and Lingshan Li},
url = {http://dl.acm.org/citation.cfm?id=2888619.2888999},
isbn = {978-1-4673-9741-4},
year = {2015},
date = {2015-12-01},
booktitle = {Proceedings of the 2015 Winter Simulation Conference},
pages = {3124–3125},
publisher = {IEEE},
address = {Huntington Beach, CA},
abstract = {Many agent simulations involve computational models of intelligent human behavior. In a variety of cases, these behavior models should be high-fidelity to provide the required realism and credibility. Cognitive architectures may assist the generation of such high-fidelity models as they specify the fixed structure underlying an intelligent cognitive system that does not change over time and across domains. Existing symbolic architectures, such as Soar and ACT-R, have been used in this way, but here the focus is on a new architecture, Sigma (!), that leverages probabilistic graphical models towards a uniform grand unification of not only the traditional cognitive capabilities but also key non-cognitive aspects, and which thus yields unique opportunities for construction of new kinds of non-modular high-fidelity behavior models. Here, we briefly introduce Sigma along with two disparate proof-of-concept virtual humans – one conversational and the other a pair of ambulatory agents – that demonstrate its diverse capabilities.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Many agent simulations involve computational models of intelligent human behavior. In a variety of cases, these behavior models should be high-fidelity to provide the required realism and credibility. Cognitive architectures may assist the generation of such high-fidelity models as they specify the fixed structure underlying an intelligent cognitive system that does not change over time and across domains. Existing symbolic architectures, such as Soar and ACT-R, have been used in this way, but here the focus is on a new architecture, Sigma (!), that leverages probabilistic graphical models towards a uniform grand unification of not only the traditional cognitive capabilities but also key non-cognitive aspects, and which thus yields unique opportunities for construction of new kinds of non-modular high-fidelity behavior models. Here, we briefly introduce Sigma along with two disparate proof-of-concept virtual humans – one conversational and the other a pair of ambulatory agents – that demonstrate its diverse capabilities.
28.
Ustun, Volkan; Rosenbloom, Paul S.
Towards Adaptive, Interactive Virtual Humans in Sigma Proceedings Article
In: Intelligent Virtual Agents, pp. 98 –108, Springer, Delft, Netherlands, 2015, ISBN: 978-3-319-21995-0.
@inproceedings{ustun_towards_2015,
title = {Towards Adaptive, Interactive Virtual Humans in Sigma},
author = {Volkan Ustun and Paul S. Rosenbloom},
url = {http://ict.usc.edu/pubs/Towards%20Adaptive,%20Interactive%20Virtual%20Humans%20in%20Sigma.pdf},
doi = {10.1007/978-3-319-21996-7_10},
isbn = {978-3-319-21995-0},
year = {2015},
date = {2015-08-01},
booktitle = {Intelligent Virtual Agents},
volume = {9238},
pages = {98 –108},
publisher = {Springer},
address = {Delft, Netherlands},
abstract = {Sigma is a nascent cognitive architecture/system that combines concepts from graphical models with traditional symbolic architectures. Here an initial Sigma-based virtual human (VH) is introduced that combines probabilistic reasoning, rule-based decision-making, Theory of Mind, Simultaneous Localization and Mapping and reinforcement learning in a unified manner. This non-modular unification of diverse cognitive, robotic and VH capabilities provides an important first step towards fully adaptive and interactive VHs in Sigma.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Sigma is a nascent cognitive architecture/system that combines concepts from graphical models with traditional symbolic architectures. Here an initial Sigma-based virtual human (VH) is introduced that combines probabilistic reasoning, rule-based decision-making, Theory of Mind, Simultaneous Localization and Mapping and reinforcement learning in a unified manner. This non-modular unification of diverse cognitive, robotic and VH capabilities provides an important first step towards fully adaptive and interactive VHs in Sigma.
29.
Rosenbloom, Paul S.; Gratch, Jonathan; Ustun, Volkan
Towards Emotion in Sigma: From Appraisal to Attention Proceedings Article
In: Proceedings of AGI 2015, pp. 142 – 151, Springer International Publishing, Berlin, Germany, 2015.
@inproceedings{rosenbloom_towards_2015,
title = {Towards Emotion in Sigma: From Appraisal to Attention},
author = {Paul S. Rosenbloom and Jonathan Gratch and Volkan Ustun},
url = {http://ict.usc.edu/pubs/Towards%20Emotion%20in%20Sigma%20-%20From%20Appraisal%20to%20Attention.pdf},
year = {2015},
date = {2015-07-01},
booktitle = {Proceedings of AGI 2015},
volume = {9205},
pages = {142 – 151},
publisher = {Springer International Publishing},
address = {Berlin, Germany},
abstract = {A first step is taken towards incorporating emotional processing into Sigma, a cognitive architecture that is grounded in graphical models, with the addition of appraisal variables for expectedness and desirability plus their initial implications for attention at two levels of the control hierarchy. The results leverage many of Sigma's existing capabilities but with a few key additions.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
A first step is taken towards incorporating emotional processing into Sigma, a cognitive architecture that is grounded in graphical models, with the addition of appraisal variables for expectedness and desirability plus their initial implications for attention at two levels of the control hierarchy. The results leverage many of Sigma's existing capabilities but with a few key additions.
30.
Kommers, Cody; Ustun, Volkan; Demski, Abram; Rosenbloom, Paul
Hierarchical Reasoning with Distributed Vector Representations Proceedings Article
In: Proceedings of 37th Annual Conference of the Cognitive Science Society, Cognitive Science Society, Pasadena, CA, 2015.
@inproceedings{kommers_hierarchical_2015,
title = {Hierarchical Reasoning with Distributed Vector Representations},
author = {Cody Kommers and Volkan Ustun and Abram Demski and Paul Rosenbloom},
url = {http://ict.usc.edu/pubs/Hierarchical%20Reasoning%20with%20Distributed%20Vector%20Representations.pdf},
year = {2015},
date = {2015-07-01},
booktitle = {Proceedings of 37th Annual Conference of the Cognitive Science Society},
publisher = {Cognitive Science Society},
address = {Pasadena, CA},
abstract = {We demonstrate that distributed vector representations are capable of hierarchical reasoning by summing sets of vectors representing hyponyms (subordinate concepts) to yield a vector that resembles the associated hypernym (superordinate concept). These distributed vector representations constitute a potentially neurally plausible model while demonstrating a high level of performance in many different cognitive tasks. Experiments were run using DVRS, a word embedding system designed for the Sigma cognitive architecture, and Word2Vec, a state-of-the-art word embedding system. These results contribute to a growing body of work demonstrating the various tasks on which distributed vector representations perform competently.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We demonstrate that distributed vector representations are capable of hierarchical reasoning by summing sets of vectors representing hyponyms (subordinate concepts) to yield a vector that resembles the associated hypernym (superordinate concept). These distributed vector representations constitute a potentially neurally plausible model while demonstrating a high level of performance in many different cognitive tasks. Experiments were run using DVRS, a word embedding system designed for the Sigma cognitive architecture, and Word2Vec, a state-of-the-art word embedding system. These results contribute to a growing body of work demonstrating the various tasks on which distributed vector representations perform competently.
31.
Garten, Justin; Sagae, Kenji; Ustun, Volkan; Dehghani, Morteza
Combining Distributed Vector Representations for Words Proceedings Article
In: Proceedings of NAACL-HLT 2015, pp. 95–101, Association for Computational Linguistics, Denver, Colorado, 2015.
@inproceedings{garten_combining_2015,
title = {Combining Distributed Vector Representations for Words},
author = {Justin Garten and Kenji Sagae and Volkan Ustun and Morteza Dehghani},
url = {http://ict.usc.edu/pubs/Combining%20Distributed%20Vector%20Representations%20for%20Words.pdf},
year = {2015},
date = {2015-06-01},
booktitle = {Proceedings of NAACL-HLT 2015},
pages = {95–101},
publisher = {Association for Computational Linguistics},
address = {Denver, Colorado},
abstract = {Recent interest in distributed vector representations for words has resulted in an increased diversity of approaches, each with strengths and weaknesses. We demonstrate how diverse vector representations may be inexpensively composed into hybrid representations, effectively leveraging strengths of individual components, as evidenced by substantial improvements on a standard word analogy task. We further compare these results over different sizes of training sets and find these advantages are more pronounced when training data is limited. Finally, we explore the relative impacts of the differences in the learning methods themselves and the size of the contexts they access.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Recent interest in distributed vector representations for words has resulted in an increased diversity of approaches, each with strengths and weaknesses. We demonstrate how diverse vector representations may be inexpensively composed into hybrid representations, effectively leveraging strengths of individual components, as evidenced by substantial improvements on a standard word analogy task. We further compare these results over different sizes of training sets and find these advantages are more pronounced when training data is limited. Finally, we explore the relative impacts of the differences in the learning methods themselves and the size of the contexts they access.
32.
Rosenbloom, Paul S.; Demski, Abram; Ustun, Volkan
Efficient message computation in Sigma’s graphical architecture Journal Article
In: Biologically Inspired Cognitive Architectures, vol. 11, pp. 1–9, 2015, ISSN: 2212683X.
@article{rosenbloom_efficient_2015,
title = {Efficient message computation in Sigma’s graphical architecture},
author = {Paul S. Rosenbloom and Abram Demski and Volkan Ustun},
url = {http://linkinghub.elsevier.com/retrieve/pii/S2212683X14000723},
doi = {10.1016/j.bica.2014.11.009},
issn = {2212683X},
year = {2015},
date = {2015-01-01},
journal = {Biologically Inspired Cognitive Architectures},
volume = {11},
pages = {1–9},
abstract = {Human cognition runs at ∼50 ms per cognitive cycle, implying that any biologically inspired cognitive architecture that strives for real-time performance needs to be able to run at this speed. Sigma is a cognitive architecture built upon graphical models – a broadly applicable state-of-the-art formalism for implementing cognitive capabilities – that are solved via message passing (with complex messages based on n-dimensional piecewise-linear functions). Earlier work explored optimizations to Sigma that reduced by an order of magnitude the number of messages sent per cycle. Here, optimizations are introduced that reduce by an order of magnitude the average time required per message sent.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Human cognition runs at ∼50 ms per cognitive cycle, implying that any biologically inspired cognitive architecture that strives for real-time performance needs to be able to run at this speed. Sigma is a cognitive architecture built upon graphical models – a broadly applicable state-of-the-art formalism for implementing cognitive capabilities – that are solved via message passing (with complex messages based on n-dimensional piecewise-linear functions). Earlier work explored optimizations to Sigma that reduced by an order of magnitude the number of messages sent per cycle. Here, optimizations are introduced that reduce by an order of magnitude the average time required per message sent.
33.
Joshi, Himanshu; Rosenbloom, Paul S.; Ustun, Volkan
Isolated word recognition in the Sigma cognitive architecture Journal Article
In: Biologically Inspired Cognitive Architectures, vol. 10, pp. 1–9, 2014, ISSN: 2212683X.
@article{joshi_isolated_2014,
title = {Isolated word recognition in the Sigma cognitive architecture},
author = {Himanshu Joshi and Paul S. Rosenbloom and Volkan Ustun},
url = {http://linkinghub.elsevier.com/retrieve/pii/S2212683X14000644},
doi = {10.1016/j.bica.2014.11.001},
issn = {2212683X},
year = {2014},
date = {2014-10-01},
journal = {Biologically Inspired Cognitive Architectures},
volume = {10},
pages = {1–9},
abstract = {Symbolic architectures are effective at complex cognitive reasoning, but typically are incapable of important forms of sub-cognitive processing – such as perception – without distinct modules connected to them via low-bandwidth interfaces. Neural architectures, in contrast, may be quite effective at the latter, but typically struggle with the former. Sigma has been designed to leverage the state-of-the-art hybrid (discrete + continuous) mixed (symbolic + probabilistic) capability of graphical models to provide in a uniform non-modular fashion effective forms of, and integration across, both cognitive and sub-cognitive behavior. Here it is shown that Sigma is not only capable of performing a simple variant of speech recognition via the same knowledge structures and reasoning algorithm used for cognitive processing, but also of leveraging its existing knowledge templates and learning algorithm to acquire automatically most of the structures and parameters needed for this recognition activity.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Symbolic architectures are effective at complex cognitive reasoning, but typically are incapable of important forms of sub-cognitive processing – such as perception – without distinct modules connected to them via low-bandwidth interfaces. Neural architectures, in contrast, may be quite effective at the latter, but typically struggle with the former. Sigma has been designed to leverage the state-of-the-art hybrid (discrete + continuous) mixed (symbolic + probabilistic) capability of graphical models to provide in a uniform non-modular fashion effective forms of, and integration across, both cognitive and sub-cognitive behavior. Here it is shown that Sigma is not only capable of performing a simple variant of speech recognition via the same knowledge structures and reasoning algorithm used for cognitive processing, but also of leveraging its existing knowledge templates and learning algorithm to acquire automatically most of the structures and parameters needed for this recognition activity.
34.
Ustun, Volkan; Rosenbloom, Paul S.; Sagae, Kenji; Demski, Abram
Distributed Vector Representations of Words in the Sigma Cognitive Architecture Proceedings Article
In: Proceedings of the 7th Conference on Artificial General Intelligence 2014, Québec City, Canada, 2014.
@inproceedings{ustun_distributed_2014,
title = {Distributed Vector Representations of Words in the Sigma Cognitive Architecture},
author = {Volkan Ustun and Paul S. Rosenbloom and Kenji Sagae and Abram Demski},
url = {http://ict.usc.edu/pubs/Distributed%20Vector%20Representations%20of%20Words%20in%20the%20Sigma%20Cognitive%20Architecture.pdf},
year = {2014},
date = {2014-08-01},
booktitle = {Proceedings of the 7th Conference on Artificial General Intelligence 2014},
address = {Québec City, Canada},
abstract = {Recently reported results with distributed-vector word representations in natural language processing make them appealing for incorporation into a general cognitive architecture like Sigma. This paper describes a new algorithm for learning such word representations from large, shallow information resources, and how this algorithm can be implemented via small modifications to Sigma. The effectiveness and speed of the algorithm are evaluated via a comparison of an external simulation of it with state-of-the-art algorithms. The results from more limited experiments with Sigma are also promising, but more work is required for it to reach the effectiveness and speed of the simulation.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Recently reported results with distributed-vector word representations in natural language processing make them appealing for incorporation into a general cognitive architecture like Sigma. This paper describes a new algorithm for learning such word representations from large, shallow information resources, and how this algorithm can be implemented via small modifications to Sigma. The effectiveness and speed of the algorithm are evaluated via a comparison of an external simulation of it with state-of-the-art algorithms. The results from more limited experiments with Sigma are also promising, but more work is required for it to reach the effectiveness and speed of the simulation.
35.
Rosenbloom, Paul S.; Demski, Abram; Han, Teawon; Ustun, Volkan
Learning via Gradient Descent in Sigma Proceedings Article
In: International Conference on Cognitive Modeling, Ottawa, Canada, 2013.
@inproceedings{rosenbloom_learning_2013,
title = {Learning via Gradient Descent in Sigma},
author = {Paul S. Rosenbloom and Abram Demski and Teawon Han and Volkan Ustun},
url = {http://ict.usc.edu/pubs/Learning%20via%20Gradient%20Descent%20in%20Sigma.pdf},
year = {2013},
date = {2013-07-01},
booktitle = {International Conference on Cognitive Modeling},
address = {Ottawa, Canada},
abstract = {Integrating a gradient-descent learning mechanism at the core of the graphical models upon which the Sigma cognitive architecture/system is built yields learning behaviors that span important forms of both procedural learning (e.g., action and reinforcement learning) and declarative learning (e.g., supervised and unsupervised concept formation), plus several additional forms of learning (e.g., distribution tracking and map learning) relevant to cognitive systems/modeling. The core result presented here is this breadth of cognitive learning behaviors that is producible in this uniform manner.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Integrating a gradient-descent learning mechanism at the core of the graphical models upon which the Sigma cognitive architecture/system is built yields learning behaviors that span important forms of both procedural learning (e.g., action and reinforcement learning) and declarative learning (e.g., supervised and unsupervised concept formation), plus several additional forms of learning (e.g., distribution tracking and map learning) relevant to cognitive systems/modeling. The core result presented here is this breadth of cognitive learning behaviors that is producible in this uniform manner.
36.
Hartholt, Arno; McCullough, Kyle; Mozgai, Sharon; Ustun, Volkan; Gordon, Andrew S
Introducing RIDE: Lowering the Barrier of Entry to Simulation and Training through the Rapid Integration & Development Environment Journal Article
In: pp. 11, 0000.
@article{hartholt_introducing_nodate-1,
title = {Introducing RIDE: Lowering the Barrier of Entry to Simulation and Training through the Rapid Integration & Development Environment},
author = {Arno Hartholt and Kyle McCullough and Sharon Mozgai and Volkan Ustun and Andrew S Gordon},
pages = {11},
abstract = {This paper describes the design, development, and philosophy of the Rapid Integration & Development Environment (RIDE). RIDE is a simulation platform that unites many Department of Defense (DoD) and Army simulation efforts to provide an accelerated development foundation and prototyping sandbox that provides direct benefit to the U.S. Army’s Synthetic Training Environment (STE) as well as the larger DoD and Army simulation communities. RIDE integrates a range of capabilities, including One World Terrain, Non-Player Character AI behaviors, xAPI logging, multiplayer networking, scenario creation, destructibility, machine learning approaches, and multi-platform support. The goal of RIDE is to create a simple, drag-and-drop development environment usable by people across all technical levels. RIDE leverages robust game engine technology while designed to be agnostic to any specific game or simulation engine. It provides decision makers with the tools needed to better define requirements and identify potential solutions in much less time and at much reduced costs. RIDE is available through Government Purpose Rights. We aim for RIDE to lower the barrier of entry to research and development efforts within the simulation community in order to reduce required time and effort for simulation and training prototyping. This paper provides an overview of our objective, overall approach, and next steps, in pursuit of these goals.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper describes the design, development, and philosophy of the Rapid Integration & Development Environment (RIDE). RIDE is a simulation platform that unites many Department of Defense (DoD) and Army simulation efforts to provide an accelerated development foundation and prototyping sandbox that provides direct benefit to the U.S. Army’s Synthetic Training Environment (STE) as well as the larger DoD and Army simulation communities. RIDE integrates a range of capabilities, including One World Terrain, Non-Player Character AI behaviors, xAPI logging, multiplayer networking, scenario creation, destructibility, machine learning approaches, and multi-platform support. The goal of RIDE is to create a simple, drag-and-drop development environment usable by people across all technical levels. RIDE leverages robust game engine technology while designed to be agnostic to any specific game or simulation engine. It provides decision makers with the tools needed to better define requirements and identify potential solutions in much less time and at much reduced costs. RIDE is available through Government Purpose Rights. We aim for RIDE to lower the barrier of entry to research and development efforts within the simulation community in order to reduce required time and effort for simulation and training prototyping. This paper provides an overview of our objective, overall approach, and next steps, in pursuit of these goals.
37.
Hartholt, Arno; McCullough, Kyle; Mozgai, Sharon; Ustun, Volkan; Gordon, Andrew S
Introducing RIDE: Lowering the Barrier of Entry to Simulation and Training through the Rapid Integration & Development Environment Journal Article
In: pp. 11, 0000.
@article{hartholt_introducing_nodate,
title = {Introducing RIDE: Lowering the Barrier of Entry to Simulation and Training through the Rapid Integration & Development Environment},
author = {Arno Hartholt and Kyle McCullough and Sharon Mozgai and Volkan Ustun and Andrew S Gordon},
pages = {11},
abstract = {This paper describes the design, development, and philosophy of the Rapid Integration & Development Environment (RIDE). RIDE is a simulation platform that unites many Department of Defense (DoD) and Army simulation efforts to provide an accelerated development foundation and prototyping sandbox that provides direct benefit to the U.S. Army’s Synthetic Training Environment (STE) as well as the larger DoD and Army simulation communities. RIDE integrates a range of capabilities, including One World Terrain, Non-Player Character AI behaviors, xAPI logging, multiplayer networking, scenario creation, destructibility, machine learning approaches, and multi-platform support. The goal of RIDE is to create a simple, drag-and-drop development environment usable by people across all technical levels. RIDE leverages robust game engine technology while designed to be agnostic to any specific game or simulation engine. It provides decision makers with the tools needed to better define requirements and identify potential solutions in much less time and at much reduced costs. RIDE is available through Government Purpose Rights. We aim for RIDE to lower the barrier of entry to research and development efforts within the simulation community in order to reduce required time and effort for simulation and training prototyping. This paper provides an overview of our objective, overall approach, and next steps, in pursuit of these goals.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper describes the design, development, and philosophy of the Rapid Integration & Development Environment (RIDE). RIDE is a simulation platform that unites many Department of Defense (DoD) and Army simulation efforts to provide an accelerated development foundation and prototyping sandbox that provides direct benefit to the U.S. Army’s Synthetic Training Environment (STE) as well as the larger DoD and Army simulation communities. RIDE integrates a range of capabilities, including One World Terrain, Non-Player Character AI behaviors, xAPI logging, multiplayer networking, scenario creation, destructibility, machine learning approaches, and multi-platform support. The goal of RIDE is to create a simple, drag-and-drop development environment usable by people across all technical levels. RIDE leverages robust game engine technology while designed to be agnostic to any specific game or simulation engine. It provides decision makers with the tools needed to better define requirements and identify potential solutions in much less time and at much reduced costs. RIDE is available through Government Purpose Rights. We aim for RIDE to lower the barrier of entry to research and development efforts within the simulation community in order to reduce required time and effort for simulation and training prototyping. This paper provides an overview of our objective, overall approach, and next steps, in pursuit of these goals.
Filter
Sorry, no publications matched your criteria.