The Accelerated Numerics group consists of a number of excellent PhD students
that I supervise, who look at various aspects of accelerating numeric problems,
in particular computational finance.
- Aryan Tavakkoli (2012-now)
- A graduate of our very own ADIC Masters programme,
Aryan is currently looking at very low-level optimisation of FPGA-accelerated financial simulations: how can we use
LUTs and FFs to efficiently implement high-level simulations of financial instruments directly, without
needing to build floating-point units?
- Peter Ogden (2012-now)
- Having worked on super-secret cyber-security at Detica, Peter is now thinking
about different ways of protecting networks, and how we need to change algorithms
and architectures to analyse THz rate channels using devices clocked in the GHz range.
- Gordon Inggs (2011-now)
- Fresh from a Masters in EE from Cape Town University, Gordon is interested in
making use of complex heterogenous systems, with a diverse combination of compute
devices and application work-loads. To explore his ideas he is using financial
computing as a case study, and trying to create prototype systems which can
statically and dynamically optimise the system for throughput and power consumption.
- Qiwei Jin (2008-now)
- Starting with acceleration of Binomial Trees for his final year project at Imperial,
Qiwei has examined many different aspects of accelerated financial computing, from
highly optimised finite-difference methods, to sophisticated forward rate modelling
in Monte-Carlo simulations.
Brahim Betkaoiu (2009-2013)
Processing irregular data-sets is used in many fields, such as protein networks,
brain-machine interfaces, and social network analysis, but can be slow in traditional
cached architectures due to the irregular memory accesses required. Brahim's work
has explored ways of maximising RAM bandwidth utilisation, by instantiating hundreds
of small specialised graph processors in an FPGA. This has provided significant
speed-ups for applications such as Graphlet counting, and using a single FPGA system
is able to beat clusters of PCs in the Graph 500 benchmark.
Anson Tse (2008-2012)
His PhD thesis encompassed quadrature methods for option pricing, control
variate methods to improve Monte-Carlo convergence, and system level methods
for balancing computation between nodes in an accelerated cluster.