@Nikol 

This works well :

master := Lowess(<$ (5 .. 256)>, convert(qw(1, 52 - 45 .. 258), Vector), fitorder = 2, bandwidth = 0.2);

NonlinearFit(B*master(t/a), <($10..235)>, wer, t, initialvalues = [a=1], output = [parametervalues]):

@tomleslie 

Thanks for the reply. I'll upload the sheet later. But you're wrong. Lowes can be used in NOnLinearfit. I get a great fit in the one-dimensional case. like B*f(A*x) where f(x):=Lowess()

@vv Thanks!

Why does this happen?

@epostma 

Thank you very match

@Markiyan Hirnyk 

Is it a bug?

@vv 

Thank you very much

@Markiyan Hirnyk 

Maple disappoints me/

rsolve({f(n)=0.5*f(n-1)+0.5*f(n+1), f(0)=1, f(a)=0},f(n));

with parameter is it impossible?

I think it's just a defect.

it's so obvious problem

I have

rsolve({ f(n+1) = f(n)+c,f(0) = 0}, {f});

return {f(n) = c (n + 1) - c}

It's a TRUE