Bio-inspired computing is a field devoted to tackling complex problems using computational methods modeled after design principles encountered in nature. The goal is to produce informatics tools with enhanced robustness, scalability, flexibility and which can interface more effectively with humans. It is a multi-disciplinary field strongly based on Biology, Computer Science, Informatics, Cognitive Science, and robotics.
Regan Predictive Algorithm (RPA)
Dr Regan’s approach in establishing a prediction result for the S&P 500 index is based on a bio-inspired methodology. The technique used by the RPA to attain a prediction result is similar to the function of the human eye. In assessing the human eye, the retina has clear inputs (light as a stream of images projected onto the retina) and outputs (optic nerve impulses), which hence leads to a well defined, unidirectional information flow. Similarly, with this concept in mind, the RPA inputs vast amounts of data into a regression core, which then selects and sorts the most relevant data based on its mean squared error (MSE) and forms the basis for the prediction result. The RPA will focus on ‘hot spots’, or particular areas of interest that are likely to influence or are influenced by the S&P 500 in the same way a human eye will react to, and process a dominant area of activity, even if it is in the peripheral vision.
Professor Toumazou of Imperial College London and Dr Regan were exchanging ideas of how best to devise a new predictive algorithm. During our conversation he told me about some novel research that had recently been undertaken on the design of an artificial retina. It was during this discussion that they both realised the parallels between the way the retina of the human eye functions and the way Dr Regan’s algorithm to predict the S&P 500 needed to operate. The human vision system is capable of processing an enormous amount of visual data that it receives from the retina. The brain reviews all of this data but can still rapidly home in on a particular area of interest; this ability being called foveation.
Through extensive optimisation and refinement, the RPA now calculates a daily predic-
tion result in ~7 hours, a speed up of over 100x from initial results.
Over an extending period of actual trading, the RPA produced a successful directional trend result of 55.1%.
World Academy of Science, Engineering and Technology International Journal of Economics and Management Engineering Vol:1, No:8, 2014 “Novel GPU Approach in Predicting the Directional Trend of the S&P 500”