How do you practically assess the contribution of a "specific" input to the NS? - page 12

New comment
 

And now look, you give the NS a static SZ during training. Okay, the network is trained.

But during operation, what will you feed it????

 
nikelodeon:

Give it up.... First of all you can't get the patterns out of NS, and secondly they (patterns) will always be different. Which will work now and which will not in the future. The NS is a very fragile tool, and to understand what the NS can and can't do is sometimes the most important thing. Because when you rely on it, you get disappointed and don't understand why it doesn't work. Because it is not the golden key and the NS is very limited in many issues.....

I have no idea how to use the NS

I'm trying to figure out if there aren't any statistical characteristics that precede the NC. Made a binary model breakout. Prediction is fine but no faith. Trying to find an NS answer but you've nixed that attempt

 
nikelodeon:

And now look, you give the NS a static SZ during training. Okay, the network is trained.

But during operation, what will you feed it????

Yes, didn't take into account, and this is the main thing, that the ZZ will have to be accompanied by something, and if they were able to, the ZZ is not needed.
 
faa1947:

Trying to find the NS answer, but you have foiled the attempt


Saves time.....
 
Swetten:
Tell me Svetlana, have you had any success using NSH???? So, just wondering.....
 
nikelodeon:
Tell me Svetlana, have you already achieved success in using NSH???? So, just curious.....

There has been some success.

If everything goes well - on the 16th of January we will start a new PAMM.

This time it will be a real live one.

Not counting all the controllable stuff.

:)

 
Well, good for you.....
 
Reshetov:
There is a very consistent pattern: ZZ peaks alternate - ..., high, then low, then high, then ... etc. etc.

...with a swift jack.

:)

 
nikelodeon:
Well, good for you.....
Thank you.
 
Swetten:

There has been some success.

If everything goes well - on the 16th of January we will start a new PAMM.

This time it will be a real live one.

Not counting all the controllable stuff.

:)


if you use the nhs, you'll probably lose in the end... but it will be interesting to watch...
1...567891011121314
New comment