Proposal: ma'o + PA

From Lojban
Revision as of 21:43, 14 July 2016 by Krtisfranks (talk | contribs) (Chunked)
Jump to navigation Jump to search

This is a brief proposal concerning the interpretation of constructs of form "ma'o PA* te'u", particularly as constant functions.

Summary

Proposal: "ma'o PA* te'u" is a grammatically valid construction which refers to a constant function, the domain set of which is implicit or specified elsewhere, such that the image of the domain set under this function is the singleton set which contains exactly the number referred to by the interpreted PA* string.

Introduction/Discussion

According to the CLL: "There is a potential semantic ambiguity in “ma'o fy. [te'u]” if “fy.” is already in use as a variable: it comes to mean “the function whose value is always ‘f’”.".

The CLL goes on to say that in practice this will not arise frequently. I am not sure of that claim, but it is explicitly addressed (in my opinion appropriately), so it is all good.

One thing which is not explicitly addressed, but which is hinted at, is the outcome of a "ma'o PA" construction. Even without the aforementioned rule for "ma'o BY" in the case of BY being a defined number (in the CLL, called a "variable" - terminology of dubious merit, but that is of no matter here), one would think that the result should be a constant function taking the value of PA once the latter is interpreted as a number. That is indeed my proposal here. The domain set should be specified in the same way that it would be specified for any function in mekso, particularly those which are expressed via "ma'o".

Examples, Justification, and Words of Caution

For example: "ma'o re te'u" is the function which is defined on some set such that it identically evaluates to 2 (id est: no matter what element of the domain set is submitted, the output is 2). Technically, outside of this domain set, the function is undefined, but it can be easily extended as necessary so as to remain constant over its newly extended domain set and, in particular, attaining only the value 2.

A generic constant function can be specified via "ma'o cy te'u" (where "cy" refers to some element of a known set of numbers) or, perhaps even more preferred, "ma'o xo'e(i) te'u".

Caution must be maintained, however. This would mean that "ma'o te'o te'u" is the constant function which evaluates to the number on its domain set. It would NOT be the exponential function with natural base, commonly denoted by "exp" or "". Likewise, "ma'o pa te'u" is the constant function which evaluates to the number 1 on its domain set while "ma'o no te'u" is the constant function which evaluates to the nunber 0 on its domain set, rather than identity functions (such as the multiplicative linear transformation which is represented by the identity matrix), although they can be closely related if not identical in some mathematical structures. I think that this is an okay sacrifice to make because it keeps the interpretation consistent and predictable for all PA. It is also utile. Moreover, it seems like a natural pair for the rule mentioned in the CLL - one which would be guessed even if the CLL rule were not prescribed. Finally, each of these alternative options ('exp', identity functions, etc.) deserve either their own words (such as "te'o'a" for 'exp'), can be built from existing words (something similar to "te'o te'a" would do for 'exp', modulo some nuance and grammatical issues), or could be referenced by a word which generically means "identity element in structure ___" (which should have at least one word); in other words, the functionality can and will be supported in other ways.

Being able to easily convert to constant functions is nice and should technically be done quite often. For example, the derivative of the number 2 does not exist because the differentiation operator is defined on a subset of functions (say, differentiable and real-valued ones) but the number 2 is just a number and not a function. But if we convert, then it is okay: "li sa'o ma'o re du li ma'o no". (Note both "ma'o"'s).

Generalizations

If a constant nonscalar tensor (or the representation thereof), for example, is converted via "ma'o", should the output be a nonscalar tensor-valued function such that it attains the original tensor on its entire domain set?

Contributors

This article was written by the following people. All ideas and opinions are theirs. First-person personal pronouns refer to them.

  • lai .krtisfranks.