top of page
W háttér 1.png

RESEARCH

Industry 4.0

mat, self d, robot.png

for
mathematical optimizations, self driving, robotics and other technologies

  

 

We have developed a novel non-parametric method for black box optimization where the target problem does not have to be differentiable and may be noisy, still our algorithm can almost always converge toward a global optimum without getting stuck at local optima. 
  

We offer this technology to help generate high quality and high quantity training data within a simulated environment to achieve the best possible model capability.

  

  

Extremely robust.    

What is the innovation?

  • General black-box optimizer

    

  • Non-parametric

  

  • (Almost) only global optima

    

  • More precision

     

  • No restart trick

     

  • No assumptions needed about problem

  

  • ​​Not required to be differentiable

   

  • Solve noisy data as well

Exceptionally good fault tolerance

kör ábra SZ.png

We've been stress testing our solution and comparing it against other well known methods with great success.    

Extremely robust.    

Here is an example of the complexity of functions it can process:

x30 - 68 - 47 * cos( x12 ) ^ x30 + 53 * 22 - x5 + 54 - 52 + x30 + 101 + x25 + 54 * x3 * sinh( x30 ) + 75 - x13 ^ 14 * tanh ( x30 ) ^ x6 - 46 * x30 ^ 81 * 55 + 97 + x30 + 15 - 47 * x10 - x11 * x20 * x26 * x30 -  sinh( x4 ) + x2 - 2 - sinh( x30 ) - 82 * tan( x15 ) * 21 + 5 * 75 * x23 - x30 * cos( x30 ) - 41 + 25 + 12 * 38 + sinh( x17 ) - 75 * 23 * x29 ^ x22 - tan( x7 ) ^ x19 - sinh( x30 ) * x8 * x30 + x18 - 3 - 18 * tan( x30 ) + 64 - 97 * x21 * x30 ^ 53 - sinh( x30 ) - 27 * 50 + 23 + 30 + 24 + 2 * 83 - 73 * 45 - 80 + 89 * x10 + 15 - x4 * tan( x28 ) * 10 + 55 * 10 * exp( x2 ) - 4 + x9 + cosh( x14 ) ^ x24 * 80 + x1 ^ sin( x27 ) ^ x16 * x4 + 49 + 26 = 100

 

( error 0, time 2.21s )

x1 = 0.43847643622855487

x2  = 0.46929961542364773 

x3 = -0.22739104890647494

x4 = 0.5419443297863478

x5 = -0.5487106409433273

x6 = 0.6862159575182608

x7 = 0.3809093553736538

x8 = -0.2790904768113744

x9 = 0.7098757786923025

x10 = 0.07842643908661656

x11 = -0.0817442135773003

x12 = 0.442795812830219

x13 = 0.38427484689331487

x14 = -0.6565246308968888

x15 = -0.8437142735692804

x16 = 0.12579756898717992

x17 = 0.4923442585462639

x18 = 0.523572942374683

x19 = 0.21694979463611894

x20 = -0.09079774813410064  

x21 = 0.519963214975034

x22 = 0.6087180159291697

x23 = 0.8651529146552689

x24 = 0.52404542812148

x25 = 0.6183467400711733

x26 = -0.44056998043975537

x27 = 0.17027625860936058

x28 = -0.13984902234792382

x29 = 0.16878526770059354

x30 = 0.1762607949000223    

How did we get to this result?

Our technological research and development of inventory optimization needed help in general problem-solving. This is how this solution was created.

   

The development has taken 2 years.

bottom of page