In this thesis topic we would like to explore a biologically inspired extension
to simple genetic programming principles (already biologically inspired), and
subsequently qualitatively and quantitatively assess the performance
characteristics of the design against the simple GP benchmark model. The models
will be generated as security portfolios on stock market price data and aim
specifically at maximising the Sharpe ratio (our measure of fitness) of the
individual modelled portfolios. The Sharpe ratio is an effectively gauge of
return accounting for risk and measures the efficiency of a portfolio of risk
bearing assets.

The motivation behind this topic is the gene-centric view of evolution
extensively developed by Richard Dawkins in The Selfish Gene and The Extended
Phenotype. Evolution has traditionally considered individual organisms as the
units of selection in its organism-centric point of view, through which the
genes that manifest into the organism's phenotype are selected for or against
on a high level. In contrast, the gene-centric point of view regards the gene
as the basic unit upon which selection acts. Genes operate both in competition
and synergy with its immediate population of other genes, and determines its
overall fitness from both this genetic environment on different levels and the
phenotypic interaction of the host organism with the host's macro-environment (Dawkins, 1976).

Portfolio optimisation is a difficult problem and current approaches do not
scale well, therefore probabilistic approximations may be useful.
Additionally, financial data is rich and readily accessible.