Machine learning in trading: theory, models, practice and algo-trading - page 2238

 
Aleksey Vyazmikin:

What do you use now?

Once you've played with the nonsense, you'll see for yourself...

no one is interested in ready-made solutions here ...

 
Fast235:

Transmit to the tablet an image from a normal computer, a mouse radio from the computer + keyboard

Put this scheme on your head ))

Tablet teeth, radio keyboard under the side and the mouse in the butt. + add remote connection lags

and enjoy your evening with a cup of coffee

 
mytarmailS:

If you play with this nonsense, you'll see for yourself...

No one is interested in ready-made solutions here...

I have a lot of ideas, too much time will pass until I implement them all, and by that time there will be new ones...

Visualization of models shows that I can predict the probability of events, but the profit/loss ratio changes with time, what I don't understand yet. It seems that in training it is necessary to split profits (green curve) and losses (red curve) more strongly sideways.


 
Maxim Dmitrievsky:

put a scheme like this on your head ))

tablet teeth, radio keyboard under the side and mouse in the butt. + add remote connection lags

and enjoy your evening with a cup of coffee

scheme works,

lag within the wifi network? they are almost none on the Internet, even in good solutions, what you intputlag, play FPS lying around?

You'll need a mouse for your laptop, and a keyboard at your side, why not?

a tablet, spin it however you want, it's lighter and touch-sensitive.

You don't need to come up with a ready-made solution, use rdp)

 
Aleksey Vyazmikin:

I have a lot of ideas, too much time will pass until I implement all of them, and by that time there will be new ones...

Visualization of models shows that the probability of events can be predicted, but the profit/loss ratio changes with time, which I do not understand yet. In my training it seems that profits (green curve) and losses (red curve) should be more separated.

The market and its participants and their algorithms change over time. It is strange that you expect a stable system trained once. Re-train once a week or every day (on trees it's fast).

 

The same problem. I have to put all the graphical interpretations of kotir into 50x50 pixels images)) making a smaller image makes no sense, it greatly reduces accuracy. A larger image will require an increase in the width of the convolution layer, and the network depth in general, which will lead to a strong increase in training time. Here too, I'm saving for a new machine... trying to curb learning on GPU at the same time.

Only I do not quite understand why to complicate the task with pictures, when you can use 1d convolution? :) picture doesn't really add any information to the row

 
Fast235:

The scheme works,

lag within the wifi network? they are almost none on the Internet, even in normal solutions, what do you intputlag, play FPS lying down?

You'll need a mouse for your laptop, and a keyboard at your side, why not?

a tablet, spin it however you want, it's lighter and touch-sensitive.

don't go for a ready-made solution now, use rdp)

thanks a lot, i'd rather use a desktop than a couch.

 
Maxim Dmitrievsky:

Thanks, I'd rather sit at a stationary than use a device on the couch.

What devices, I do not use an additional keyboard, a tablet + mouse without a cowl for quick navigation

You need to develop artificial intelligence, not the MO, to help)

 
Fast235:

What devices, I do not use a keyboard, a tablet + mouse without a pen for quick navigation

Have you tried using a normal laptop? Try it, tell me ))

 
Maxim Dmitrievsky:

Have you tried using a normal laptop? Try it, you'll tell me.)

Slim keyboard + touch, any will be flawed