Investigating scaling of an abstracted LCS utilising ternary and S-Expression alphabets

Utilising the expressive power of S-Expressions in Learning Classifier Systems often prohibitively increases the search space due to increased flexibility of the endcoding. This work shows that selection of appropriate S-Expression functions through domain knowledge improves scaling in problems, as expected. It is also known that simple alphabets perform well on relatively small sized problems in a domain, e.g. ternary alphabet in the 6, 11 and 20 bit MUX domain. Once fit ternary rules have been formed it was investigated whether higher order learning was possible and whether this staged learning facilitated selection of appropriate functions in complex alphabets, e.g. selection of S-Expression functions. This novel methodology is shown to provide compact results (135-MUX) and exhibits potential for scaling well (1034-MUX), but is only a small step towards introducing abstraction to LCS.

Generalised golden ratios over integer alphabets

It is a well known result that for β ∈ (1,1+√52) and x ∈ (0,1β−1) there exists uncountably many (ǫi)∞i=1 ∈ {0,1}N such that x = P∞i=1ǫiβ−i. When β ∈ (1+√52,2] there exists x ∈ (0,1β−1) for which there exists a unique (ǫi)∞i=1 ∈ {0,1}N such that x=P∞i=1ǫiβ−i. In this paper we consider the more general case when our sequences are elements of {0, . . . , m}N. We show that an analogue of the golden ratio exists and give an explicit formula for it.

Learning Latin the ancient way: Latin textbooks from the ancient world

A collection of Latin textbooks used by Greek speakers in the Roman empire, translated and annotated for modern readers. Includes reading material such as dialogues, phrasebooks, Colloquia of the Hermeneumata Pseudodositheana, stories about the Trojan war, Aesop's fables, legal treatises, and model letters; grammatical works from Dositheus and Charisius; glossaries/lexica; prose compostion exercises; alphabets. Some texts are transliterated.