Sam Ritchie comments

Results 351 comments of


                                            Sam Ritchie

Support all output types in reverse-mode

Qs for @littleredcomputer: - what do you think of this protocol approach? I am not sure there is a way around doing two passes over the output structure, annoyingly... but...

Support all output types in reverse-mode

Superceded by #179, #182, #183, #185

Port, optimize AD benchmarks from Siskind + Pearlmutter's work

Thanks @qobi !

Benchmark and (possibly) speed up term-rewriting

I'm currently taking this work on in sicmutils/sicmutils#155 .

Allow for equation representation (ie symbolic `=`)

I've started this over in sicmutils/sicmutils#262 - I've got `=` plugged in. Next I'll generalize to handle the other comparison operators.

Allow for equation representation (ie symbolic `=`)

@hcarvalhoalves , I have come around to the idea that we want special constructors for this — maybe `sym:=` or something, but not `=`. it would be too easy to...

Implement Brent's Method for root finding

Not presumptuous, but it is indeed more complicated! The issue is that the description of the algorithm here https://en.wikipedia.org/wiki/Brent%27s_method#Algorithm does not match the code in Brent's book, in chapter 4....

`g/partial-derivative` for structures shouldn't use fn impl for NON-structural inputs

I've thought about this more, and I see now that this will only work if you call D on a function that takes a single numerical (or symbolic) argument. The...

`g/partial-derivative` for structures shouldn't use fn impl for NON-structural inputs

This implies that maybe we should treat these cases separately, EVEN if they're both covered by `D`...

Flesh out literal fn syntax for structured args, values to match scmutils

Covering this in sicmutils/sicmutils#262 .