Let  

f := beta/(2*sigma*GAMMA(1/beta))*exp(-(abs(x-mu)/sigma)^beta);

where mu::real, beta > 0, sigma > 0.
How can we obtain the expression of F?

F := int(f, x=-infinity..s)

Reverse problem (a priori simpler):
It's known that 

F := 1/2+signum(x-mu)/(2*GAMMA(1/beta))*(GAMMA(1/beta)-GAMMA(1/beta, abs((x-mu)/sigma)^beta))

How can we check that diff(F, x)=f?
Even with the assumptions on beta, mu and sigma and additional assumptions x>0 or x<0, I can't verify that diff(F, x)=f.

 

I want to estimate numerically the value of P (that is the probability that f exceeds the value 12 when x, y and z are uniformly distributed within the box [-Pi, Pi]³).

restart
f := sin(x)+7*sin(y)^2+0.1*z^4*sin(x);
Omega := [x, y, z] =~ [(-Pi..Pi)$3]:
                                2        4       
               sin(x) + 7 sin(y)  + 0.1 z  sin(x)


h := Heaviside(f-12):
P := Int(h, Omega);

Here is a simple Monte Carlo estimation of P.

f_MC  := x -> sin(x[1])+7*sin(x[2])^2+0.1*x[3]^4*sin(x[1]); 
h_MC  := x -> Heaviside(f_MC(x) - 12);
omega := -Pi, Pi;
                         
P_MC := proc(N)
  local Z := Statistics:-Sample(Uniform(omega), [N, 3]):
  local F := Vector(N, n -> h_MC(Z[n])):
  local K := add(F):
  local P := evalf(add(F)/N):
  local q := Statistics:-Quantile(Normal(0, 1), 0.005, numeric):
  local e := -q*sqrt(P*(1-P)/N):
  printf("A bilateral 99%% confidence interval that Prob(f > 12) is %1.3e +/- %1.2e\n", P, e);
end proc:

P_MC(10^6)
A bilateral 99% confidence interval that Prob(f > 12) is 1.643e-02 +/- 3.27e-04

I was hoping to get a value for P using  evalf/Int with some method.
But I didn't succeed with any of the methods for multiple integration:

• The following command returns 0 for n=1 and 2 and Int(h, Omega) if n >=3

  evalf( Int(h, Omega, 'epsilon=1e-n', 'method=_MonteCarlo') );
  

   

• And all my attempts with methods from the Cuba library have resulted in an unevaluated Int(h, Omega) integral. 
   

Could you help me to estimate P using  evalf/Int ?
TIA

Hi,
Can anyone tell me what's going wrong with int(Heaviside(f-1/2), x=0..1, y=0..1); ?
 

restart:

f := x*y:

JH := int(Heaviside(f-1/2), x=0..1, y=0..1);  #???
                               1
                               -
                               4

 JP := int(piecewise(x*y>1/2, 1, 0), x=0..1, y=0..1); 
                          1   1      
                          - - - ln(2)
                          2   2      

# integrate "1" over the domain where f > 1/2

S := solve(f=1/2, x):
JS := int(1, x=S..1, y=1/2..1);
                          1   1      
                          - - - ln(2)
                          2   2      

Thanks in advance
 

In a french magazine written by High Schools teachers I found this problem:

let a, b, p, q four strictly positive integers such that a > b^2 and p > q+1;
find 4-tuples (a, b, p, q) such that 

(a^2 - b^4) = p!/q!

Given the source of this problem I suspect that there is a trick to answering this question.
After some hours spent, I have found no general method to solve it, only a few solutions (first one and second one are almost obvious), for instance

rel := a^2 - b^4 = p!/q!:

eval(rel, [a= 5, b=1, q=1, p=4]);   
eval(rel, [a=11, b=1, q=1, p=5]);
eval(rel, [a=71, b=1, q=1, p=7]);
eval(rel, [a= 2, b=1, q=2, p=3]);
eval(rel, [a=19, b=1, q=2, p=6]);
eval(rel, [a=21, b=3, q=2, p=6]);

Do you have any idea how to solve this problem?
Could it be handled by Maple (without a systematic exploration of a part of N^4)?

Thanks in advance

Hi,

I solve numerically an ODE system which depends on 25 parameters.
I want to know the maximum value of the time at which a specific event is triggered, when these parameters vary (independently the one of the other) within a 25 dimensional hyperbox.

To solve this maximization problem, which I assume is local¹, I would like to use NLPSolve

The attached zip (I use Mac OSX) contains:

• a file that gives a full description of the problem (the original ".pages" file and its export as pdf),
• an m file wich contains the solution procedure plus a few other variables needed to solve the problem,
• the mw file which contains:
  • the reading of the m file,
  • the procedure (OBJ) which returns the trigerring time,
  • many attemps to find its maximum value using NLPSolve.

None of my attempts at using NLPSolve gave me the expected answer

Could you help me to fix this?

Thanks in advance

mw_and_m_files.zip

1: Initializing a local method (see the explanation file) with different points gave me different optimizers but all of them led to rather close values of the objective function. Thus the problem is either global instead of local, or convergence might not be achieved for all the initialization points (but keep in mind that I'm not interested in the location of the minimizer(s) but in the maximum value of if the time when the event is triggered).