Bug in temporal.decay

[jimdale]

The temporal.decay function should presumably compute feature(t0)exp([t-t0]/tau), where t0 is the time when feature was last non-zero, but instead computes 2 ** ((-1 * feature(t) / tau_halflife).

[hhegre]

Are you sure this is a bug? It is convenient to have 2 as the base. Or what am I missing?

[jimdale]

It's not so much a bug as a function with a slightly misleading name, in that it does not compute time decays on its own - you have to chain it with time_since.

I don't like using half-lives much (I prefer e-folding times), but that is just personal preference