Hybrid neural networks. - page 2
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
You mean there are a lot of scales on the net and that's why it takes a long time to learn? How many scales? How long does it take to learn?
Oh yes, the network at first stages is full-bonded, or like convolution networks, but there are a lot of layers). And all this happiness is multiplied by 10 and starts mating. Each of them has to be processed, i.e. we have 10x. And if you have an idea to teach a profitable trick, then I have to calculate all the time interval for each generation and run it through each progeny. This operation totally killed me with its resource-intensiveness and I go back to my original question.Oh yes, the network at first stages is full-bonded, or like convolutional nets, but with many layers). So it's multiplied by 10 and starts pairing. Each of them has to be processed, i.e. we have 10x. And if you have an idea to teach a profitable trick, then I have to calculate all the time interval for each generation and run it through each progeny. This operation totally killed me with its resource-intensiveness and I go back to my original question.
Number of layers?
Number of layers?
The old fashioned [50]-60-39-2. Full-bodied.The old fashioned [50]-60-39-2. Full-blown.
And as for the genetic code, look in the private line.
Still haven't answered my question, "How many scales? How long is the training time?"
But my understanding is that there are only 10 individuals in the colony. That's very few. And you're wasting your time allowing everyone in the population to interbreed. It's not efficient.
Apparently there's something wrong with the algorithm as well, since it's taking so long to work.
I'm using a population of 200 individuals. Each individual has up to 300,000 genes. Learning takes 10 minutes.
Try running a simple function with two variables first, like this:
F=MathPow(MathCos(2*x*x)-1.1,2)+MathPow(MathSin(0.5*x)-1.2,2)-MathPow(MathCos(2*y*y)-1.1,2)+MathPow(MathSin(0.5*y)-1.2,2)
with a search range of -5 to 5. In this range of variables the function has 1 global maximum (x=-3.315699...; y=-3.072485...) and one global minimum (x=3.0702175...; y=3.3159335...)
I have a geneticist looking for the minimum in 380 milliseconds. And in the same amount of time the maximum.
Optimise the algorithm on simple functions. Then start training neural networks.
Yesterday I wrote a 10-15-10-1 grid
moving on...
But as far as I understand, there are only 10 individuals in the colony. That's very few. And you shouldn't allow everyone in the population to interbreed. It's not efficient.
I am using a population of 200 individuals. Each individual has up to 300000 genes. It takes 10 minutes to learn.
Where did I write that I have them all interbreeding? Of course not the 80%-20% corridor.
Didn't you read about XOR or something?
Disclose the structure of the grid (which is 200 specimens each).
Do you recommend increasing the population? If you don't mind, set up a small experiment. How long will it take to train a simple task (time, number of populations) for 200 individuals and for 25 individuals. Let's leave the rest unchanged. I haven't experimented at all at this point.
Optimisation parameters:
1. Continuation probability corridor 80-20%
2. weight step 0.1-0.001
3. Gene mutation probability 20-50%
Yesterday I wrote a 10-15-10-1 grid
>> go on...
nice. Two incoming bars?