ForumSections

Machine learning in trading: theory, models, practice and algo-trading - page 3276

New comment
Maxim Dmitrievsky #:
Oops, that's it
))

Greetings, engine room!

I've also created a powerful neural network.

Trained it for pounddoll and normal flight.


Alexander Ivanov #:

Greetings engine room!

I, too, have created a powerful neural network.

Trained it for pounddoll and normal flight.

Can I see the trading history so that I can see the trades?

Renat Akhtyamov #:

can I see the trading history so that I can see the trades?


I've trained on pounddoll and it's fine... But there's not much history.

But I'm glad.

Now I'm making the neural network train itself as it trades.

I mean, we should get a smart bot. And we don't have to train it every week.

Alexander Ivanov #:


I've trained on pounddoll, and it's fine. But there's not much history.

But I'm glad.

Is this a demo? (stop 5-10 pips)

Renat Akhtyamov #:

is this a demo?

Yeah.

Alexander Ivanov #:

yes

I see.

but how to train it in such a way that it would eat pips of 100 4-digit points, is it possible?

I mean, to get away from pips.
mytarmailS #:

about the strategy ansables

https://buildalpha.wordpress.com/2018/11/20/buildalpha-ensemble-strategies-reduce-overfitting-by-combining-strategies/

----------------------------------------

All we need is a metric of strategy retraining to know whether the strategy will work on new data or not, everything else is solvable....

----------------------------------------

I have an idea to take several approaches to detect overtraining, mine is based on auto.arima, Prado "PBO". I have an idea to take several approaches to detect overtraining, mine based on auto.arima, Prado "PBO", something else maybe, throw them in as predictors and teach AMO to predict the probability of overtraining and make it a metric.

Alternatively.

Extremely interesting. Overtraining is the second pillar of MO. The first is clearing predictors from the rubbish, the simplest. But there's a third pillar, looking ahead. Here there is no thinking at all.

Renat Akhtyamov #:

I see.

But how to train it so that it would eat points of 100 4 digits, is it possible?

Anything is possible.

1...326932703271327232733274327532763277327832793280328132823283...3749
New comment