Decentralised minimum-time consensus

We consider the discrete-time dynamics of a network of agents that exchange information according to a nearest-neighbour protocol under which all agents are guaranteed to reach consensus asymptotically. We present a fully decentralised algorithm that allows any agent to compute the final consensus value of the whole network in finite time using the minimum number of successive values of its own state history. We show that the minimum number of steps is related to a Jordan block decomposition of the network dynamics, and present an algorithm to compute the final consensus value in the minimum number of steps by checking a rank condition of a Hankel matrix of local observations. Furthermore, we prove that the minimum number of steps is related to graph theoretical notions that can be directly computed from the Laplacian matrix of the graph and from the minimum external equitable partition. © 2013 Elsevier Ltd. All rights reserved.

Language model combination and adaptation using weighted finite state transducers

In speech recognition systems language model (LMs) are often constructed by training and combining multiple n-gram models. They can be either used to represent different genres or tasks found in diverse text sources, or capture stochastic properties of different linguistic symbol sequences, for example, syllables and words. Unsupervised LM adaptation may also be used to further improve robustness to varying styles or tasks. When using these techniques, extensive software changes are often required. In this paper an alternative and more general approach based on weighted finite state transducers (WFSTs) is investigated for LM combination and adaptation. As it is entirely based on well-defined WFST operations, minimum change to decoding tools is needed. A wide range of LM combination configurations can be flexibly supported. An efficient on-the-fly WFST decoding algorithm is also proposed. Significant error rate gains of 7.3% relative were obtained on a state-of-the-art broadcast audio recognition task using a history dependently adapted multi-level LM modelling both syllable and word sequences. ©2010 IEEE.

Assessing seasonal precipitation trends in India using parametric and non-parametric statistical techniques

This paper provides an insight into the long-term trends of the four seasonal and annual precipitations in various climatological regions and sub-regions in India. The trends were useful to investigate whether Indian seasonal rainfall is changing in terms of magnitude or location-wise. Trends were assessed over the period of 1954-2003 using parametric ordinary least square fits and non-parametric Mann-Kendall technique. The trend significance was tested at the 95% confidence level. Apart from the trends for individual climatological regions in India and the average for the whole of India, trends were also specifically determined for the possible smaller geographical areas in order to understand how different the trends would be from the bigger spatial scales. The smaller geographical regions consist of the whole southwestern continental state of Kerala. It was shown that there are decreasing trends in the spring and monsoon rainfall and increasing trends in the autumn and winter rainfalls. These changes are not always homogeneous over various regions, even in the very short scales implying a careful regional analysis would be necessary for drawing conclusions regarding agro-ecological or other local projects requiring change in rainfall information. Furthermore, the differences between the trend magnitudes and directions from the two different methods are significantly small and fall well within the significance limit for all the cases investigated in Indian regions (except where noted). © 2010 Springer-Verlag.

Effective handling of dialogue state in the hidden information state POMDP-based dialogue manager

Effective dialogue management is critically dependent on the information that is encoded in the dialogue state. In order to deploy reinforcement learning for policy optimization, dialogue must be modeled as a Markov Decision Process. This requires that the dialogue statemust encode all relevent information obtained during the dialogue prior to that state. This can be achieved by combining the user goal, the dialogue history, and the last user action to form the dialogue state. In addition, to gain robustness to input errors, dialogue must be modeled as a Partially Observable Markov Decision Process (POMDP) and hence, a distribution over all possible states must be maintained at every dialogue turn. This poses a potential computational limitation since there can be a very large number of dialogue states. The Hidden Information State model provides a principled way of ensuring tractability in a POMDP-based dialogue model. The key feature of this model is the grouping of user goals into partitions that are dynamically built during the dialogue. In this article, we extend this model further to incorporate the notion of complements. This allows for a more complex user goal to be represented, and it enables an effective pruning technique to be implemented that preserves the overall system performance within a limited computational resource more effectively than existing approaches. © 2011 ACM.