Date: Thursday 1 September
Who: Brendon Lovett
Seminar type: Tutorial Seminar
Time: 4-5pm
Where: Conference Room
Outline:
Electronic and Optical Properties of Semiconductors
Date:
Part I - August 26, 2005, 12 Midday, Interaction Room
Part II - August 30, 2005, 4pm, Conference Room, Physics Annexe
Who: Sanjeev Naguleswaran
Seminar type: Research seminar
Abstract:
Current Automated Planning techniques to extract valid plans from a plan specification are typically exponential in complexity. In addition, representation of planning problems as Markov Decision Processes (MDP) and their subsequent solution suffers from the “state space explosion” problem where the number of states grows exponentially large with the size of the problem.
This presentation will endeavor to describe research undertaken with the purpose of developing a reduced complexity/tractable automated planning methodology. This work is based on concepts borrowed from fields such as Mathematics, Physics, Computer Science and Engineering.
The representation of the planning problem as a directed bi-partite graph which can be unfolded to a tree like structure will be described first. The equipping of this structure with probabilities and the Markov property will then be briefly explored resulting in a Markov Decision Process with a reduced number of states. A method for adapting the Grover Search Algorithm for use within a Dynamic Programming algorithm such as Value Iteration is then described as a possible low complexity solution to optimise the MDP.
Preliminary investigations into adapting quantum walks to extract a valid plan from a graphical structure will also be presented.
Date: August 5, 2005
Who: Terry Rudolph
Seminar type: Informal chalk and talk
When: 12 Midday
Where: Interaction Room, Physics Annexe
Abstract:
The primary technological hurdles facing linear optical quantum computing are in the construction of single photon sources and detectors that lie under a fault tolerant threshold. Broadly speaking, two possible approaches to this problem suggest itself. The first is to work experimental physicists a lot harder, give them more money to buy better grad students, and to sit and wait until they deliver sources and detectors precise at the 0.01% level. Skeptics of such an approach ever delivering suitable devices, which include most experimental physicists I know, advocate the second approach: work theoretical physicists a lot harder, give them more money to buy better grad students, and sit and wait until they deliver more relaxed fault tolerant thresholds. This informal chalk-and-talk will outline some of my recent results in the latter of these two approaches. Specifically I’ll discuss how we can get by with single photon sources which are only 66% pure if we have perfect single photon detectors; alternatively we can get by with detectors that are only 50% efficient if we have perfect sources. The tradeoff between these two thresholds (i.e. imperfect sources and imperfect detectors) is still open, primarily because everywhere I go people make me give different talks - despite the well known fact that my talks, as this one will definitively prove, have vanishing trace distance from the maximally mixed state…