Parallel Evaluation of Numerical Models for Algorithmic Trading
| Thesis title in Czech: | Parallel Evaluation of Numerical Models for Algorithmic Trading |
|---|---|
| Thesis title in English: | Parallel Evaluation of Numerical Models for Algorithmic Trading |
| Key words: | paralelizace, GPU, Xeon Phi, algoritmické obchodování, support vector machines |
| English key words: | parallelization, GPU, Xeon Phi, algorithmic trading, support vector machines |
| Academic year of topic announcement: | 2014/2015 |
| Thesis type: | diploma thesis |
| Thesis language: | angličtina |
| Department: | Department of Software Engineering (32-KSI) |
| Supervisor: | doc. RNDr. Martin Kruliš, Ph.D. |
| Author: | hidden - assigned and confirmed by the Study Dept. |
| Date of registration: | 31.10.2014 |
| Date of assignment: | 31.10.2014 |
| Confirmed by Study dept. on: | 20.11.2014 |
| Date and time of defence: | 05.09.2016 09:30 |
| Date of electronic submission: | 28.07.2016 |
| Date of submission of printed version: | 28.07.2016 |
| Date of proceeded defence: | 05.09.2016 |
| Opponents: | RNDr. Filip Zavoral, Ph.D. |
| Guidelines |
| Algorithmic trading on the stock market has become a serious business for many companies and corporations. Algorithmic trading applies speculative approach to stock market transactions, where the decision to buy or sell individual trading instruments is left to a sophisticated mathematical model. The trading algorithm is capable of performing many transactions in short periods of time, thus creating significant profit on relatively small fluctuations of prices.
Very fast evaluation of the mathematical model and thus a low latency response to the market behaviour is the key feature of a successful high-frequency trading algorithm. The mathematical model offers many opportunities for concurrent evaluation; however, ensuring low latency processing requires extremely precise load balancing. Furthermore, the initial work decomposition and subsequent synchronization and result aggregation must be implemented in a special manner to minimize the serial overhead of the algorithm. The main objective of this work is to analyze the existing models for algorithmic trading from the perspective of parallel processing with a particular focus on parallel accelerators such as GPUs and Xeon Phi devices. The analysis will help design a prototype implementation of the trading model which intensively utilizes these parallel architectures. The performance of the prototype implementation will be experimentally evaluated and compared to a baseline serial and multicore CPU algorithm. |
| References |
| David B. Kirk, Wen-mei W. Hwu: Programming Massively Parallel Processors, Second Edition: A Hands-on Approach, 2012, ISBN: 0124159923
Jason Sanders, Edward Kandrot: CUDA by Example: An Introduction to General-Purpose GPU Programming, NVIDIA 2010, ISBN: 0-13-138768-5 Matthew Scarpino: OpenCL in Action: How to Accelerate Graphics and Computations, Manning Publications 2011, ISBN: 1617290173 James Jeffers, James Reinders: Intel Xeon Phi Coprocessor High-Performance Programming, 2013, ISBN: 978-0124104143 Cristianini, Nello and Shawe-Taylor, John: An Introduction to Support Vector Machines and other kernel-based learning methods, Cambridge University Press, 2000. ISBN 0-521-78019-5 Catanzaro, Bryan and Sundaram, Narayanan and Keutzer, Kurt: Fast Support Vector Machine Training and Classification on Graphics Processors, in International Conference on Machine Learning, 2008 |