This is partially in response to a question I posted earlier, although the problem in this case appears to be distinct from the problem there, so I am temporarily filing this as a separate question.


Length[Reduce[Abs[c1] + Abs[c2] + Abs[c3] + Abs[c4] <= 19, {c1, c2, c3, c4}, Integers]]

yields 97281, which is in full agreement with the result obtained when Reduce is replaced by Solve.

However, executing

Length[Reduce[Abs[c1] + Abs[c2] + Abs[c3] + Abs[c4] <= 22, {c1, c2, c3, c4}, Integers]]

yields 42736. This is obviously impossible, since the number of $\mathbb{Z}^4$ lattice points within a taxicab ball is a strictly increasing function of radius.

Does anyone know what is going wrong here? I am going to temporarily solve the problem via alternate, more delicate approaches seeing as both Solve[] and Reduce[] have failed to give correct answers for $r\geq20$, but any insight as to this sudden behavioral discontinuity would be appreciated.

EDIT: It may be possible that the lower than expected result from Length may be due to Reduce outputting a solution set whose expression depth was not just that of the simple Or[...] structure that it normally gives for smaller values of $r$. If anyone can confirm this, I'll close this question.

  • $\begingroup$ I upvoted your last question. But this time, I really think it's your job to find out for certain whether the output actually is incorrect, rather than that it just has an unexpected structure. $\endgroup$ – Oleksandr R. Dec 25 '13 at 2:23
  • $\begingroup$ At the very least, it contains this (call the result sols and then run this): Cases[sols, Except@HoldPattern@And[__Equal], {1}]. Is this right or wrong? I'm not sure. As this is also the last element, I think Reduce is just telling us that it gave up exhaustively enumerating solutions at that point. $\endgroup$ – Oleksandr R. Dec 25 '13 at 2:32
  • $\begingroup$ @OleksandrR.: Artes recently told me about the option "ReduceOptions" -> "ExhaustiveSearchMaxPoints", which I'll try fiddling with once I'm finished with dinner. $\endgroup$ – DumpsterDoofus Dec 25 '13 at 2:36

If you try:

ListLinePlot[{#, Length@Reduce[Abs[c1] + Abs[c2] + Abs[c3] + Abs[c4] <= # && 
                 And @@ Thread[{c1, c2, c3, c4} >= 0], {c1, c2, c3, c4}, Integers]} & /@ 
                                                                            Range[2, 20]]

You'll get:

Mathematica graphics

It only means that Reduce[] is expressing its results in a different way. Take a look at them and then post your own answer :)

  • 1
    $\begingroup$ Thanks, I looked at them and it's clear that it's decided to stop using exhaustive enumeration. $\endgroup$ – DumpsterDoofus Dec 25 '13 at 4:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.