Algorithm Optimisation Championship. - page 39
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Funny video. It's a pity they didn't show what the algorithm would do if they started the evolution from a flat spot on the surface - always give the algorithm a chance to catch the rise. And there are only two parameters.
Maybe the author of the video is among us, speak back.
Funny video. It's a pity they didn't show what the algorithm would do if they started the evolution from a flat spot on the surface - always give the algorithm a chance to catch a lift. And there are only two parameters.
That's what it looks like:
//---
Andrey Dik:
Maybe the author of the video is among us, speak up.
The author writes that if you have any questions, you can try to contact him. ;)
Randy Olson:
Please feel free to direct them here to ask questions - or email me: http://randalolson.com/contact
That's a cool idea with the video. We should do something like that.
There would be nothing, the grey dots would change chaotically and be evenly spaced across the surface.
This would be bad behaviour for the algorithm. A better one would be an accelerated propagation to all areas. The more often the value of the FF does not change, the more likely it is that the max has already been reached and it may be just a local extremum. Therefore it is necessary to start investigating unknown regions as soon as possible. It should look as if the points are scattering faster and faster at each frame. It appears that the algorithm will spread out on flat surfaces, and on steep slopes it will start to climb faster and faster. Uniform propagation should be only on monotonically varying FF values over the entire definition area (wave).
The directions for the run-up are not definable, so there is only chaotic wandering.
The directions for the run-up are not definable, so only chaotic rambling remains.
Determined - in either direction from the parent. And the longer there is no improvement in the parent, the faster the offspring must scatter to the sides.
If improvement occurs - on the contrary, the offspring appear close to the parent, i.e. direction - towards the parent.
There are always two directions - from parent and towards parent. Depending on dynamics of change of FF values one or another direction should be chosen.
But for the author of the video species always "hang out" not far from the parent, and the unexplored areas remain unexamined.
The algorithm is likely to converge very quickly on continuous functions and fail on complex discrete ones with sharp peaks. And even surfaces, judging by video algorithm is difficult.
If you know English enough to have a casual small talk, please contact the author, invite him to the contest.
Determined - in either direction from the parent. And the longer there is no improvement in the parents, the faster the offspring must scatter to the sides.
If improvement occurs, it is the opposite, the offspring appearing close to the parent.
There are always two directions - away from the parent and towards the parent. Depending on dynamics of change of FF values one or another direction should be chosen.
But at the author of the video the individuals "hang out" always near the parent, and the unexplored areas remain unexplored.
The algorithm is likely to converge very quickly on continuous functions and fail on complex discrete ones with sharp peaks. And even surfaces, judging by video algorithm is difficult.
If anyone knows English to a decent degree to have a casual small talk, please contact the author, invite him to the champ.
Well, I know English to a decent degree. I'll invite him in.