トップ   一覧 単語検索 最終更新   ヘルプ   最終更新のRSS

***** 専攻談話会(セミナー) [#t67d9871]

集中講義などでいらっしゃる先生など、お心あたりのある方は専攻幹事 [[岸本:http://www.is.titech.ac.jp/~kishi/]] (kishimoto at is.titech.ac.jp) までお知らせください。
集中講義などでいらっしゃる先生など、お心あたりのある方は専攻幹事 [[Tuan:http://www.is.titech.ac.jp/~tuan/]] (tuan at is.titech.ac.jp) までお知らせください。


**2013年度第1回 [#gdaa1d28]

日時: ''2013年4月18日(木) 13:20-14:50''

会場: ''西8号館W棟8階W809号室''

講演者: '' Anne Bouillard 氏''

題目: ''Exact Worst-case Delay for FIFO-multiplexing Tandems''

This talk is about computing the actual worst-case end-to end
delay for a flow in a tandem of network under some general assumptions
using the network calculus theory, that is originally based on the
(min,plus) algebra. Unfortunately, using those algebraic methods lead to
very pessimistic bounds. We will see that we can indeed model this
network as a mixed interger linear program and drastically improve those

**2012年度第4回 [#j4f0fd9f]

日時: ''2013年3月21日(木) 15:00-16:30''

会場: ''西8号館E棟10階E1011号室''

講演者: '' Martin Mu"ller 氏''

題目: ''Move Quality in Monte Carlo Simulation:
A Case Study using the Fuego Go Program''

Despite half a dozen years of intense research, designing effective
simulation policies for Monte Carlo Tree search in Go is still
considered something of a black art, and driven largely by trial and
error. Important ideas that have evolved include pattern- and
tactics-based playouts, simulation balancing, and several schemes to
dynamically modify simulation policies online. In this study, we take
an in-depth look at what happens when the Go program Fuego runs its
playouts. We develop several methods to evaluate the quality of moves
played in these simulations, and we evaluate the contribution of the
different components of Fuego's playout policy. We study the
distribution of both the number of blunders, or result-changing moves,
and the absolute loss - in terms of number of points - for many
variations of the Fuego playout policy. We use this study to identify
an improvement to the Fuego default policy.

(Joint work with Sumudu Fernando)

**2012年度第3回 [#j4f0fd9f]

日時: ''2013年3月8日(金) 16:00-17:00''

会場: ''西8号館W棟10階1008号室''

講演者: ''Benjamin Burton 氏(The University of Queensland)''

題目: ''Untangling knots using combinatorial optimisation''

Unknot recognition is the algorithmic problem of determining whether a
knot (i.e., a closed loop) in 3-dimensional space can be untangled.
It is a major unsolved problem as to whether this problem has a fast
polynomial time solution.  Here we present the first conclusive
algorithm for unknot recognition which, although exponential time in
theory, exhibits a clear polynomial time behaviour in exhaustive
practical experiments.  The algorithm makes significant use of
techniques from combinatorial optimisation, and uses a
branch-and-bound framework with linear programming steps.  We also
introduce the topological software package Regina (in which the
algorithm was implemented), and discuss the relevance of our results
to other difficult problems in computational topology.

This is joint work with Melih Ozlen.

講演後には Burton 氏との懇親会を予定しています。奮ってご参加ください。

**2012年度第2回 [#j4f0fd9f]

日時: ''2012年11月12日(月) 15:00-16:30''

会場: ''西8号館E棟10階1011号室''

講演者: ''S.R.K. Branavan 氏''

題目: ''Grounding Linguistic Analysis in Control Applications''

Natural languages are the medium in which the majority of humanity's collective knowledge is recorded and communicated.  If machines were able to automatically access and leverage this information, they could effectively perform many tasks that are currently considered to be computationally intractable, thus requiring human involvement.  Today, the only way to infuse human knowledge into computational algorithms is to have humans in the loop - i.e., to manually encode the knowledge into heuristics, through annotations, or directly into the model structure itself.

Our ultimate goal is to automate this process, so that machines can access required knowledge directly from text.  One path to this goal is to perform a semantic interpretation of text by grounding the textual information in the objects, actions and dynamics of the physical world.  From a linguistic viewpoint, this grounding of language in control applications presents a very natural notion of language semantics.  I.e., by allowing us to define semantics with respect to the control application, and avoid imposing subjective human notions of correctness.

In this talk, I will explore two aspects of the connection between language and control applications: first, how the semantic analysis of language can be driven by control performance; and second, how information from text can be leveraged to improve performance in complex control systems.  These two aspects are in fact complementary, and language analysis is central to them both.  Addressing these aspects jointly is particularly challenging.  However, as our empirical results show, connecting the semantic analysis of language to control applications allow us to leverage the synergy between the two to simultaneously learn both language analysis and application control with little or no prior knowledge.

**2012年度第1回 [#j4f0fd9f]

日時: ''2012年4月19日(木) 13:20-14:50''

会場: ''西8号館W棟8階W809号室''

講演者: ''Tuan Phung-Duc 氏 (東京工業大学 数理・計算科学専攻)''

題目: ''Markov Chain and Queue with State-dependent Transition Rates''

In this talk, we consider continuous time Markov chains with
state-dependent transition rates arising from either retrial queues or
infinite server queues. First, for multiserver retrial queues, we show
that in the case of three and four servers, the joint stationary
distribution of the number of busy servers and the number of retrying
customers can be expressed in terms of continued fractions.
Furthermore, we formulate the general case with any number of servers
by a level-dependent QBD process for which a memory saving algorithm
based on matrix continued fractions is presented. Next, we briefly
present a topic on infinite server queues with synchronized
abandonment, where we develop a new methodology to obtain either
closed form solution or recursion for the stationary distribution of
the number of busy servers.

**過去の談話会 [#db336cc6]