44,715 Hits in 6.4 sec

More Efficient Exploration with Symbolic Priors on Action Sequence Equivalences [article]

Toby Johnstone, Nathan Grinsztajn, Johan Ferret, Philippe Preux
2021 arXiv   pre-print
In this paper, we consider the problem of exploiting priors about action sequence equivalence: that is, when different sequences of actions produce the same effect.  ...  Incorporating prior knowledge in reinforcement learning algorithms is mainly an open question.  ...  For both environments, a greater equivalence set leads to a more efficient exploration.  ... 
arXiv:2110.10632v2 fatcat:jg6vugxkijfvhbczttdvrvieoq

SDRL: Interpretable and Data-efficient Deep Reinforcement Learning Leveraging Symbolic Planning [article]

Daoming Lyu, Fangkai Yang, Bo Liu, Steven Gustafson
2019 arXiv   pre-print
Experimental results validate the interpretability of subtasks, along with improved data efficiency compared with state-of-the-art approaches.  ...  The task-level interpretability is enabled by relating symbolic actions to options.This framework features a planner -- controller -- meta-controller architecture, which takes charge of subtask scheduling  ...  A planner uses prior symbolic knowledge to perform longterm planning by a sequence of symbolic actions (subtasks) that achieve its intrinsic goal; 2.  ... 
arXiv:1811.00090v4 fatcat:wzp64myruzgsnducsqrkzrtwtm

Parsing Videos of Actions with Segmental Grammars

Hamed Pirsiavash, Deva Ramanan
2014 2014 IEEE Conference on Computer Vision and Pattern Recognition  
with variable durations and orderings.  ...  Real-world videos of human activities exhibit temporal structure at various scales; long videos are typically composed out of multiple action instances, where each instance is itself composed of sub-actions  ...  Note that α is symbol-specific x m rather than rulespecific r m ; this allows different rules to share data models for equivalent symbols.  ... 
doi:10.1109/cvpr.2014.85 dblp:conf/cvpr/PirsiavashR14 fatcat:z4un5zgej5gstl6hqe7uzlnau4

Symbolic Input-Output Conformance Checking for Model-Based Mutation Testing

Bernhard K. Aichernig, Martin Tappler
2016 Electronical Notes in Theoretical Computer Science  
In this approach, a possibly non-deterministic action system model is used as basis for generating a number of mutants.  ...  This paper presents an approach to use symbolic input output conformance checking for mutation-based test case generation.  ...  They explore the defined product graph "on the fly" and return a diagnostic sequence of actions leading to a state, where non-conformance may be observed, if the checked IOLTSs are not ioco-conform.  ... 
doi:10.1016/j.entcs.2016.01.002 fatcat:32577qon5bcv3bvrfm7rgshfqi

SDRL: Interpretable and Data-Efficient Deep Reinforcement Learning Leveraging Symbolic Planning

Daoming Lyu, Fangkai Yang, Bo Liu, Steven Gustafson
Experimental results validate the interpretability of subtasks, along with improved data efficiency compared with state-of-the-art approaches.  ...  The task-level interpretability is enabled by relating symbolic actions to options.This framework features a planner – controller – meta-controller architecture, which takes charge of subtask scheduling  ...  Finally, recent work on integrating symbolic planning with RL (Yang et al. 2018; Lu et al. 2018) provides SP+RL frameworks where symbolic plans based on prior knowledge can guide RL for meaningful exploration  ... 
doi:10.1609/aaai.v33i01.33012970 fatcat:tbj2xd55wvhmdhrmqtnqmgnd4u

Systematically Exploring the Behavior of Control Programs

Jason Croft, Ratul Mahajan, Matthew Caesar, Madan Musuvathi
2015 USENIX Annual Technical Conference  
We develop a new technique to systematically explore the behavior of control programs.  ...  Because control programs depend intimately on absolute and relative timing of inputs, a key challenge that we face is to systematically handle time.  ...  To address this challenge, we build on prior work and combine symbolic execution with model checking [6] .  ... 
dblp:conf/usenix/CroftMCM15 fatcat:xbdfsty46nbnfp2eyhzm2vkyze

Reinforcement Learning with Information-Theoretic Actuation [article]

Elliot Catt, Marcus Hutter, Joel Veness
2021 arXiv   pre-print
In this work we explore and formalize a contrasting view, namely that actions are best thought of as the output of a sequence of internal choices with respect to an action model.  ...  This view is particularly well-suited for leveraging the recent advances in large sequence models as prior knowledge for multi-task reinforcement learning problems.  ...  Although in the case of uninformative action binarization one can map a discounted external setup to an equivalent discounted internal setup (Majeed and Hutter 2020) , attempt-ing a more general construction  ... 
arXiv:2109.15147v1 fatcat:dsl67a42szf6xlhy3ezapp5mcm

Autonomously constructing hierarchical task networks for planning and human-robot collaboration

Bradley Hayes, Brian Scassellati
2016 2016 IEEE International Conference on Robotics and Automation (ICRA)  
As the presented method relies on the structure of the task itself, our work imposes no particular type of symbolic insight into motor primitives or environmental representation, making it applicable to  ...  Similarly, if two or more skills perform equivalent functions from a goal fulfillment perspective, they must be classified as instantiations of the same skill (e.g., pressing a button with the top or side  ...  A chain has a starting vertex with outdegree one and a termination vertex with in-degree one.  ... 
doi:10.1109/icra.2016.7487760 dblp:conf/icra/HayesS16 fatcat:4hvpcix6o5daxpur4jgkgbl2ky

Automata Learning with On-the-Fly Direct Hypothesis Construction [chapter]

Maik Merten, Falk Howar, Bernhard Steffen, Tiziana Margaria
2012 Communications in Computer and Information Science  
Our DHC algorithm starts with a one-state hypothesis that it successively extends using a direct construction approach.  ...  the automata construction process: the learning algorithm continues to complete its hypothesis, providing intuition to a field of formal methods otherwise dominated by algorithms that largely operate on  ...  Automata learning can reveal such properties with little or no prior knowledge on the inner workings of exploration targets.  ... 
doi:10.1007/978-3-642-34781-8_19 fatcat:ypmdcdhhrnhydiee4piu3asqm4

PLANS: Neuro-Symbolic Program Learning from Videos

Raphaël Dang-Nhu
2020 Neural Information Processing Systems  
Rule-based approaches offer correctness guarantees in an unsupervised way as they inherently capture logical rules, while neural models are more realistically scalable to raw, high-dimensional input, and  ...  Recent years have seen the rise of statistical program learning based on neural models as an alternative to traditional rule-based systems for programming by example.  ...  first exploring likely solutions.Ellis et al. (2018) learn a bias optimal strategy based on Levin search(Levin, 1973), with the goal of efficiently allocating compute time to different parts of the program  ... 
dblp:conf/nips/Dang-Nhu20 fatcat:yua7txrzpzgo5jvvotdi7me7pi

Learning from Humans [chapter]

Aude G. Billard, Sylvain Calinon, Rüdiger Dillmann
2016 Springer Handbook of Robotics  
We close with a look on the use of language to guide teaching and a list of open issues.  ...  We emphasize the importance of choosing well the interface and the channels used to convey the demonstrations, with an eye on interfaces providing force control and force feedback.  ...  One common way is to segment and encode the task according to sequences of predefined actions, described symbolically.  ... 
doi:10.1007/978-3-319-32552-1_74 fatcat:wtcftkgkwveexpfbmnkcebi5wu

Transfer learning across heterogeneous robots with action sequence mapping

Balaji Lakshmanan, Ravindran Balaraman
2010 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems  
We extend the framework to enable the robot to learn an equivalence between certain sequences of its actions and certain sequences of actions of the other robot.  ...  Transfer learning refers to reusing the knowledge gained while solving a task, to solve a related task more efficiently.  ...  It can also be extended to be used with more information, say multiple trajectories, partial action equivalence or partial map of the environment etc.  ... 
doi:10.1109/iros.2010.5649422 dblp:conf/iros/LakshmananR10 fatcat:siwb5jylqrh4vlggcsfhgybeae

TDM: Trustworthy Decision-Making Via Interpretability Enhancement

Daoming Lyu, Fangkai Yang, Hugh Kwon, Wen Dong, Levent Yilmaz, Bo Liu
2021 IEEE Transactions on Emerging Topics in Computational Intelligence  
Moreover, a TDM-based algorithm is introduced to demonstrate the unification of symbolic planning with other sequential-decision making algorithms, reaping the benefits of both.  ...  Human-robot interactive decision-making is increasingly becoming ubiquitous, and trust is an influential factor in determining the reliance on autonomy.  ...  It also makes sub-policy for each subtask to be more easily learned and subtasks more easily sequenced. Experimental Results.  ... 
doi:10.1109/tetci.2021.3084290 fatcat:76olgi5wkrgnho7mcrbkg3y5mm

Symbolic Verification of Cache Side-channel Freedom [article]

Sudipta Chattopadhyay, Abhik Roychoudhury
2018 arXiv   pre-print
At the core of our framework is a novel symbolic verification technique based on automated abstraction refinement of cache semantics.  ...  The power of such a framework is to allow symbolic reasoning over counterexample traces and to combine it with runtime monitoring for eliminating cache side channels during program execution.  ...  Each synthesized patch captures a symbolic condition ν on input variables and a sequence of actions that needs to be applied when the program is processed with inputs satisfying ν .  ... 
arXiv:1807.04701v1 fatcat:cjukpybayvdcbpbf5wzypbck5y

Finding counterexamples from parsing conflicts

Chinawat Isradisaikul, Andrew C. Myers
2015 SIGPLAN notices  
Steve Blackburn and Ben Hardekopf helped with the review process. This work was supported by Office of Naval Research grant N00014-13-1-0089 and National Science Foundation grant 0964409.  ...  Each state contains a collection of transitions on symbols and a collection of production items. Each transition is either a shift action on a terminal symbol or a goto on a nonterminal symbol.  ...  A distinct sequence of parser actions taken by each copy describes one possible derivation of the counterexample.  ... 
doi:10.1145/2813885.2737961 fatcat:lgqukdm2izf3ja4rbs2tkwex3a
« Previous Showing results 1 — 15 out of 44,715 results