The Experts below are selected from a list of 360 Experts worldwide ranked by ideXlab platform
Farn Wang - One of the best experts on this subject based on the ideXlab platform.
-
RTSS - Efficient Model-Checking of Dense-Time Systems with Time-Convexity Analysis
2008 Real-Time Systems Symposium, 2008Co-Authors: Farn WangAbstract:The evaluation of successor or predecessor state spaces through time progress is a central component in the model-checking algorithms of dense-time automata. The time progress operator takes the concavity of a path condition into consideration and usually results in high complexity in the evaluation. Previous algorithms in this aspect usually assume that the location invariance condition of an automaton are convex in the dense-time state space and use a more efficient algorithm for time progress evaluation. In fact, the restriction of location invariance condition convexity can be further relaxed to that of time-convexity for a broader range of application of the more efficient algorithm. In this work, we present techniques for the efficient model-checking of dense-time automata by taking the time-convexity of path conditions into consideration. We first identify a class of TCTL formulas that only characterize time-convex state spaces. The class includes several important types of TCTL formulas, including some timed inevitabilities with deadlines. We then present a new formulation for the efficient evaluation of timed inevitabilities with time-convex path conditions. The new formulation also leads to a new technique for the approximate evaluation of timed inevitabilities with better precision. Finally, we report our implementation and experiment.
-
ATVA - Time-Progress Evaluation for Dense-Time Automata with Concave Path Conditions
Automated Technology for Verification and Analysis, 2008Co-Authors: Farn WangAbstract:The evaluation of successor or predecessor state spaces through time progress is a central component in the model-checking algorithm of dense-time automata. The definition of the time progress operator takes into consideration of the path condition of time progress and usually results in high complexity in the evaluation. Previous algorithms in this aspect usually assume that the original location invariance conditions of an automaton are convex in the dense-time state space. Based on this assumption, efficient algorithms for convex path conditions can be designed for reachability analysis. However, it is not clear whether the path conditions are still convex in the general setting of TCTL model-checking. In this work, we discuss the concept of time-convexity that allows us to relax the restrictions on the application of time-progress evaluation algorithm for convex path conditions. Then we give examples in TCTL model-checking that engenders time-concave path conditions even when the original automaton location invariance conditions are time-convex. Then we present two techniques that allow us to apply the evaluation algorithms for time-convex path conditions to time-concave path conditions. Finally, we report our experiment with the techniques. For some benchmarks, our techniques may enhance the performance of model-checking by an order of magnitude.
-
Efficient Model-Checking of Dense-Time Systems with Time-Convexity Analysis
2008 Real-Time Systems Symposium, 2008Co-Authors: Farn WangAbstract:The evaluation of successor or predecessor state spaces through time progress is a central component in the model-checking algorithms of dense-time automata. The time progress operator takes the concavity of a path condition into consideration and usually results in high complexity in the evaluation. Previous algorithms in this aspect usually assume that the location invariance condition of an automaton are convex in the dense-time state space and use a more efficient algorithm for time progress evaluation. In fact, the restriction of location invariance condition convexity can be further relaxed to that of time-convexity for a broader range of application of the more efficient algorithm. In this work, we present techniques for the efficient model-checking of dense-time automata by taking the time-convexity of path conditions into consideration. We first identify a class of TCTL formulas that only characterize time-convex state spaces. The class includes several important types of TCTL formulas, including some timed inevitabilities with deadlines. We then present a new formulation for the efficient evaluation of timed inevitabilities with time-convex path conditions. The new formulation also leads to a new technique for the approximate evaluation of timed inevitabilities with better precision. Finally, we report our implementation and experiment.
Patrick Dumberry - One of the best experts on this subject based on the ideXlab platform.
-
An Overview of State Succession Issues Arising as a cResult of an Armed Conflict
International Investment Law and the Law of Armed Conflict, 2019Co-Authors: Patrick DumberryAbstract:This chapter addresses the question of what happens to legal protections of foreign investors under a BIT when an armed conflict leads to a situation of State succession. It first examines the actual practice of States concerning the continuation of BITs. Such continuation depends on the express or tacit agreement of both States concerned, i.e. the successor State and the “other State Party” to the BIT which was signed by the predecessor State. The chapter also examines the question of what happens to BITs when the States concerned have not agreed on any such continuation. I will argue that BITs do not automatically continue to be in force as of the date of succession unless both States concerned have agreed to such a continuation.
-
State Succession to State Contracts: A New Framework of Analysis for an Unexplored Question
The journal of world investment and trade, 2018Co-Authors: Patrick DumberryAbstract:This article examines the surprisingly unexplored question of State succession to State contracts. It argues that the successor State is not bound by such contracts. The article puts forward a new framework of analysis with regards to this question emphasizing on the importance of two elements. First, the solution depends on the type of succession involved. The analysis shows that, while the principle of succession is firmly established for certain types (cession of territory, unification, integration), it is not for others (secession and dissolution). Second, the solution depends on a number of different factors and circumstances, including the existence of a ‘territorial nexus’ between a contract and the successor State, whether the contract was signed by an organ of a territorial unit of the predecessor State which has a structural continuity with the successor State, and the need to avoid any situation of unjust enrichment.
-
Is a New State Responsible for Obligations Arising from Internationally Wrongful Acts Committed Before its Independence in the Context of Secession
Canadian Yearbook of international Law Annuaire canadien de droit international, 2016Co-Authors: Patrick DumberryAbstract:This article addresses an aspect of one of the most controversial issues in the field of state succession: whether there is a succession of states to obligations arising from an internationally wrongful act committed by the predecessor state against a third state before the "date of succession." More specifically, the article addresses the issue in the context of secession: who from the continuing state or the new state (the "secessionist" successor state) should he held responsible for the obligation to repair towards the injured third state after the date of succession. There are basically two possible answers to this question: the obligation remains that of the continuing state or the obligation is transferred to the new state. This article will first critically examine the arguments advanced by scholars for whom the successor state does not take over the obligations arising from internationally wrongful acts committed by the predecessor state before the date of succession. We will offer a less dogmatic approach to the issue whereby the solution essentially depends on different factors and circumstances as well as the types of succession of states involved. This article will then examine state practice and case law in the context of secession, which shows that the continuing state usually continues its previous responsibility for internationally wrongful acts committed before the date of succession, notwithstanding the transformation affecting its territory (see, the position taken by Austria and the Allies in the context of the break-up of the Austria-Hungary Dual Monarchy; Polish and German court decisions in the context of the secession of Poland; the position adopted by East German after the Second World War; several examples in the context of the break-up of the U.S.S.R.; claims arising in the context of the secession of Belgium). A more controversial issue is whether this principle of non-succession should always apply in the context of secession and whether there should not be some circumstances that would instead call for the new state to be responsible for internationally wrongful acts committed before the date of succession. This article is an attempt to systematically establish under which specific circumstances secessionist states take over the obligation to repair towards injured third states. The author identifies five such circumstances.
-
Responsibility for Crimes Committed by the Ottoman Empire against the Armenian Population: Are the Rules of State Succession to International Responsibility of Any Use?
Hague Yearbook of International Law Annuaire de La Haye de Droit International Vol. 26 (2013), 2013Co-Authors: Patrick DumberryAbstract:This article examines the question of whether or not Turkey can be held responsible for internationally wrongful acts committed by the Ottoman Empire against the Armenian population before, during, and after the First World War. This article analyses the relevant rules of State succession to international responsibility based on the assumption that the break-up of the Ottoman Empire was a case of dissolution and that Turkey became a new State in 1923. I will consider several examples of State practice and case law in the context of dissolution of States as well as four specific circumstances where a new State should take over the obligations arising from the commission of an internationally wrongful act by the predecessor State before independence.
-
The Controversial Issue of State Succession to International Responsibility in Light of Recent State Practice
2008Co-Authors: Patrick DumberryAbstract:This paper addresses the issue of succession of States to international responsibility. In other words, does the successor State takes over obligations arising from internationally wrongful act committed by the predecessor State against a third State before the date of succession? The traditional answer given by scholars is that there is no transfer of responsibility from one State to another. This paper critically assesses the arguments used in doctrine in support of this position. The first argument is that as a State is not responsible for acts committed by other States, only the State which has actually committed an internationally wrongful act should be held responsible for it. The second argument relates to the "personal character" of an internationally wrongful act which would not enable its transfer from one State to another: action personalis moritur cum persona. The present article focuses on recent State practice of the 1990s. This paper examines several examples of State practice in the context of the integration of the German Democratic Republic in the Federal Republic of Germany, the dissolution of the Socialist Federal Republic of Yugoslavia, the break-up of the U.S.S.R., the independence of Namibia and also the ICJ Gabc"ikovo Nagymaros Project Case in the context of the dissolution of Czechoslovakia.This practice clearly shows that the strict and automatic principle of non-succession is not representative of contemporary international law. In fact, not a single recent example was found where the successor State refused to be held responsible for pre-succession obligations. However, the analysis of State practice does not show the existence of any general obligation for the successor State to succeed to the predecessor State's international responsibility.
Hermann Ney - One of the best experts on this subject based on the ideXlab platform.
-
Comparison of feedforward and recurrent neural network language models
ICASSP IEEE International Conference on Acoustics Speech and Signal Processing - Proceedings, 2013Co-Authors: Martin Sundermeyer, B. Freiberg, Ralph Schluter, Ilya Oparin, Daniel Povey, Lukas Burget, Arnab Ghoshal, Jean-luc Gauvain, Karel Vesely, Hermann NeyAbstract:Research on language modeling for speech recognition has increasingly focused on the application of neural networks. Two competing concepts have been developed: On the one hand, feedforward neural networks representing an n-gram approach, on the other hand recurrent neural networks that may learn context dependencies spanning more than a fixed number of predecessor words. To the best of our knowledge, no comparison has been carried out between feedforward and state-of-the-art recurrent networks when applied to speech recognition. This paper analyzes this aspect in detail on a well-tuned French speech recognition task. In addition, we propose a simple and efficient method to normalize language model probabilities across different vocabularies, and we show how to speed up training of recurrent neural networks by parallelization. © 2013 IEEE.
-
lstm neural networks for language modeling
Conference of the International Speech Communication Association, 2012Co-Authors: Martin Sundermeyer, Ralf Schluter, Hermann NeyAbstract:Neural networks have become increasingly popular for the task of language modeling. Whereas feed-forward networks only exploit a fixed context length to predict the next word of a sequence, conceptually, standard recurrent neural networks can take into account all of the predecessor words. On the other hand, it is well known that recurrent networks are difficult to train and therefore are unlikely to show the full potential of recurrent models. These problems are addressed by a the Long Short-Term Memory neural network architecture. In this work, we analyze this type of network on an English and a large French language modeling task. Experiments show improvements of about 8 % relative in perplexity over standard recurrent neural network LMs. In addition, we gain considerable improvements in WER on top of a state-of-the-art speech recognition system.
Raphael Deimel - One of the best experts on this subject based on the ideXlab platform.
-
IROS - Reactive Interaction Through Body Motion and the Phase-State-Machine
2019 IEEE RSJ International Conference on Intelligent Robots and Systems (IROS), 2019Co-Authors: Raphael DeimelAbstract:Between humans, body motion is an intuitive and ubiquitous means to coordinate interactions, but so far collaborative robots have not embraced this mode of communication. One reason for this is that conventional behavior generation systems use finite state machines which make it exceptionally difficult to ingest and produce the inherently continuous and concurrent information flow that body motion provides. We propose a new, reactive motion generator based on a dynamical system instead. It mimics a conventional state machine except that transitions are not instantaneous but time-extended, reversible, and have phases. Moreover, consecutive states and transitions may overlap during execution. Most notably, more than one transition can be active at the same time if they share predecessor state. Together, these unique features enable instantaneous and gradual responses to novel information and continuous decision processes without changing the state graph itself. We demonstrate the system’s capabilities in an object handover task coordinated by body motion.
-
Reactive Interaction Through Body Motion and the Phase-State-Machine
2019 IEEE RSJ International Conference on Intelligent Robots and Systems (IROS), 2019Co-Authors: Raphael DeimelAbstract:Between humans, body motion is an intuitive and ubiquitous means to coordinate interactions, but so far collaborative robots have not embraced this mode of communication. One reason for this is that conventional behavior generation systems use finite state machines which make it exceptionally difficult to ingest and produce the inherently continuous and concurrent information flow that body motion provides. We propose a new, reactive motion generator based on a dynamical system instead. It mimics a conventional state machine except that transitions are not instantaneous but time-extended, reversible, and have phases. Moreover, consecutive states and transitions may overlap during execution. Most notably, more than one transition can be active at the same time if they share predecessor state. Together, these unique features enable instantaneous and gradual responses to novel information and continuous decision processes without changing the state graph itself. We demonstrate the system's capabilities in an object handover task coordinated by body motion.
Martin Sundermeyer - One of the best experts on this subject based on the ideXlab platform.
-
Comparison of feedforward and recurrent neural network language models
ICASSP IEEE International Conference on Acoustics Speech and Signal Processing - Proceedings, 2013Co-Authors: Martin Sundermeyer, B. Freiberg, Ralph Schluter, Ilya Oparin, Daniel Povey, Lukas Burget, Arnab Ghoshal, Jean-luc Gauvain, Karel Vesely, Hermann NeyAbstract:Research on language modeling for speech recognition has increasingly focused on the application of neural networks. Two competing concepts have been developed: On the one hand, feedforward neural networks representing an n-gram approach, on the other hand recurrent neural networks that may learn context dependencies spanning more than a fixed number of predecessor words. To the best of our knowledge, no comparison has been carried out between feedforward and state-of-the-art recurrent networks when applied to speech recognition. This paper analyzes this aspect in detail on a well-tuned French speech recognition task. In addition, we propose a simple and efficient method to normalize language model probabilities across different vocabularies, and we show how to speed up training of recurrent neural networks by parallelization. © 2013 IEEE.
-
lstm neural networks for language modeling
Conference of the International Speech Communication Association, 2012Co-Authors: Martin Sundermeyer, Ralf Schluter, Hermann NeyAbstract:Neural networks have become increasingly popular for the task of language modeling. Whereas feed-forward networks only exploit a fixed context length to predict the next word of a sequence, conceptually, standard recurrent neural networks can take into account all of the predecessor words. On the other hand, it is well known that recurrent networks are difficult to train and therefore are unlikely to show the full potential of recurrent models. These problems are addressed by a the Long Short-Term Memory neural network architecture. In this work, we analyze this type of network on an English and a large French language modeling task. Experiments show improvements of about 8 % relative in perplexity over standard recurrent neural network LMs. In addition, we gain considerable improvements in WER on top of a state-of-the-art speech recognition system.