Information Technology Reference
In-Depth Information
Fig. 1.
Immune memory development [1]
apoptotic pathways and revert back to a naive like state. The key difference
to naive cells however is that these revertant cells are able to homeostatically
turnover, producing clones to sustain knowledge of an antigen experience over
the long term. These two distinct memory pools, and the transfer mechanism
between them, represent a key difference to other memory theories, and prove
the inspiration for memory development in our algorithm.
In our solution the equivalent of the short term memory pool is generated
using a derivative of the popular clonal selection algorithm [5] which proliferates
all successfully bound candidates. The short term memory pool evolves through
a special form of mutation, and is regulated through apoptosis. Successful candi-
dates from the short term memory pool then transfer to the long term memory
pool for permanent storage. This pool can then be utilised during future antigen
presentations to aid in identification. These mechanisms are discussed in detail
in Section 4.
4
An Immune Inspired Trend Evaluation and Prediction
Solution
The pseudo code for the proposed Trend Evaluation Algorithm (TEA) is detailed
in Program 1. Each of the significant operations in the TEA is then described
in the subsequent sections. All parameters noted in these sections have been
chosen using educated guesses based on previous experience, no formal sensitivity
analysis has been performed to date but will form part of our future work.
4.1
Tracker Pool Construction and Initialisation
The TEA comprises a population of individual 'trackers' whose purpose is to
identify the price trends located within an antigen. Each tracker is a vector
consisting of multiple price change estimates, much like the antigen. The price
estimates are generated using a Gaussian distribution and converted to price
