What to feed to the input of the neural network? Your ideas... - page 25
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
Optimisation of weights for 2021
Forward almost 2 years: from 2022 to 2023-10-29
Maxim discounts forwards like that too. What is the point of earning 20% for almost 2 years ?????????? Force the system up to a maximum drawdown of 10% or use depo 10%. And there you will see at once better character of the system. If of course the goal is not to surprise lamers with a trick.
Just three entrances
in[1] = Close[1] - Close[2];
in[2] = Close[2] - Close[3];
in[3] = Close[3] - Close[4];
Scale optimisation for 2021
Forward almost 2 years: from 2022 to 2023-10-29
What's the catch. - It's one of hundreds or one of thousands of sets. You'll never find it among the plum ones.
What's the good news? - Neuronics can work on any data, as long as it reflects a little bit of what's happening on the chart
Looks good. Did you run the weights in the optimiser? Only three layers in the network, taking into account the input and output neuron?
is that on the daily? - not enough trades, the result could be a false positive.
On the hourly
Looks good. Did you do that in the optimiser? Only three layers in the network, taking into account the input and output neuron?
Yeah, just a regular optimiser. 2 layers of 10 neurons with a sigmoid. And 3 inputs of increments of hourly closing prices.
The falsely successful period for the Jew is from 2021 to September 2022. After that, usually any grid pours, because it is always adjusted to the trend of 2021. But here, it went further. Moreover, the 2020 untrained - also up, not so beautiful, but it was caught.
The problem is that increasing the inputs to 4 did not give any improvement, the frequency of trades is higher, there are more sets fitted, but you can't catch a normal set like that.
Maxim discounts such forwards too. And what is the point of earning 20% for almost 2 years ??????????. Force the system up to a maximum drawdown of 10% or use depo 10%. And there you will see at once better character of the system. If of course the goal is not to surprise lamers with a trick.
Here it would be better to move from the dead point, what systems from the MO branch)
Forward almost 2 years: from 2022 to 2023-10-29
It would be nice to get off the ground here, what systems there are from the MoD branch)
Yeah, just a regular optimiser. 2 layers of 10 sigmoid neurons. And 3 inputs of increments of hourly closing prices.
The falsely successful period for eur is from 2021 to september 2022. After that, usually any grid pours, because it is always adjusted to the trend of 2021. But here, it went further. Moreover, the 2020 untrained - also up, not so beautiful, but it was caught.
The problem is that increasing the inputs to 4 did not give any improvement, the frequency of trades is higher, there are more sets fitted, but you can't catch a normal set like that.
So maybe it makes sense to train sequentially - at the first stage generate many trades near zero but with sufficient expectation for profitable ones, and at the second stage look for the network configuration that will improve the data from the first stage? So you can make several sequential networks.
So maybe it makes sense to train sequentially - at the first stage generate a lot of deals near zero but with sufficient mat expectation for profitable ones, and at the second stage search for the network configuration that will improve the data from the first stage? So you can make several sequential networks.
You say something clever, but I am a simple machine from the article here https://www.mql5.com/ru/articles/497 twisting and turning.
I can't even imagine the realisation of the task