Mole End

The Mole's grandmother was born in another country, one where English is not the native language.  We will return to this point later.  Our immediate point of departure, however, is that the Mole is currently burnishing his credentials in pursuit of a return to more technical work and has, to that end, been taking online courses on machine learning and other "data science" topics.  The first course he took was quite theoretical, abstract, and mathematical, and the normal equation method of solving a linear regression problem was presented as

(X^TX)^-1X^Ty

using the universally familiar notation of linear algebra.

As the Mole completes homework assignments he has implemented this several times, in several languages, and expects to have to do it at least once more in at least one more language.  Each reimplementation requires a wholly uninstructive detour into syntax, a cost that the Mole would justify as an investment if he could settle with some confidence on one of these languages as the one he was likely to use primarily in the near future.  But the language wars rage on -- it was always thus, and always will be -- and the poor Mole cannot predict what will be most useful.  So far, the Mole has written

pinv(X' * X) * X' * y

in Octave and

coef(lm(y~X[,2]+X[,3]))

and

solve(t(X) %*% X) %*% t(X) %*% y

in R.  He fully expects to add Python to the list very shortly.  Then what?  Julia?  Java?  Javascript?

The Mole's grandmother spoke excellent, nearly unaccented, English.  Yet traces of her native tongue remained:  when she counted silently to herself (while kntting, for example), she reverted to her first language.  So it is with the Mole, who leaves as an exercise for the reader determining why his mind first imagines this code:

((⌹((⍉X)+.×X))+.×⍉X)+.×Y