Do you think you will be able to address this issue next week or do you i need to specify further?

[nikolas-joyce]

wanted to have V1 style flexibility to setting different objective function and swarm visibility that i could not find in the current V2 base and genetic opt.

I think this involves porting the DPS alphas and the optunity notebooks used for calibration.

https://10.0.1.2:8888/notebooks/strategies/Strategy_DSP_LPBP_Combination_StrategyModule_new%20%2B%20Optunity%20smart%20optimization.ipynb

this notebook, which Dmitry build for the various versions of the production alphas, has a few useful features that i wanted to highlight.

1. the ability to optimize based on a selection of different objective functions.
   2. ability to alter the size of the in and out of sample data usage.

3.

functionally formatted solutions that can easily be copied into another notebook for deployment.( this feature may be redundant with your v2 alpha.save function)

4) the ability to examine visually the resulting sub swarm and picked equity curve.

[alexveden]

this notebook, which Dmitry build for the various versions of the production alphas, has a few useful features that i wanted to highlight.

the ability to optimize based on a selection of different objective functions.

Objective functions engine is implemented, but now we have only 'net profit' and 'sharpe' obj. functions int he V2 framework.

ability to alter the size of the in and out of sample data usage.

Already implemented, you can set size of in-sample and out-of-sample window for Walk-forward optimization.

functionally formatted solutions that can easily be copied into another notebook for deployment.( this feature may be redundant with your v2 alpha.save function)

I'm not sure, because you will have different active swarm set on each WFO step, however it could be possible in case of hardcoded rules / opts.

he ability to examine visually the resulting sub swarm and picked equity curve.

The swarm picking algorithm designed and works not like V1 swarms, because I had the machine learning and GA models in my mind. It's still possible to implement full swarm calculation and plot, but I'm not sure about the way of visualizing the picked and average swarm.

I believe that I can create notebook for full swarm visualization, and the average swarm, where 'picked' swarm will be the equity of alpha itself.

Also I'm not clear, are you wanted to implement Optunity optimization in the V2, or we are talking about some useful features of these notebooks?

[nikolas-joyce]

I think I need v1 style swarm picking to be able to have flexibility I require. I think but I am not sure that you would need to use the same optunity logic to give the different features from the outlined notebooks. It seems that the v1 and v2 selection algorithm is sufficiently different that we cannot have the same functionality in the genetic algorithm for reasons you have pointed out. Therefore I would like you to mimic the v1 selection in v2 so I can have the same notebook functionality for swarm construction.

Thx

Sent from my iPhone

On Jul 31, 2017, at 6:48 AM, alexveden notifications@github.com<mailto:notifications@github.com> wrote:

this notebook, which Dmitry build for the various versions of the production alphas, has a few useful features that i wanted to highlight.

the ability to optimize based on a selection of different objective functions.

Objective functions engine is implemented, but now we have only 'net profit' and 'sharpe' obj. functions int he V2 framework.

ability to alter the size of the in and out of sample data usage.

Already implemented, you can set size of in-sample and out-of-sample window for Walk-forward optimization.

functionally formatted solutions that can easily be copied into another notebook for deployment.( this feature may be redundant with your v2 alpha.save function)

I'm not sure, because you will have different active swarm set on each WFO step, however it could be possible in case of hardcoded rules / opts.

he ability to examine visually the resulting sub swarm and picked equity curve.

The swarm picking algorithm designed and works not like V1 swarms, because I had the machine learning and GA models in my mind. It's still possible to implement full swarm calculation and plot, but I'm not sure about the way of visualizing the picked and average swarm.

I believe that I can create notebook for full swarm visualization, and the average swarm, where 'picked' swarm will be the equity of alpha itself.

Also I'm not clear, are you wanted to implement Optunity optimization in the V2, or we are talking about some useful features of these notebooks?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/trendmanagement/Tmqr-framework-2/issues/31#issuecomment-319072192, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ARobLMz2wPTfhNvypBVe1GT3f7NgcDKzks5sTds8gaJpZM4OluFw.

[nikolas-joyce]

Even in the v2 how you restrict the swarm can be quite powerful. Essentially I need an efficient way to uncover how to restrict the calibration for different outcomes.

Sent from my iPhone

On Jul 31, 2017, at 6:48 AM, alexveden notifications@github.com wrote:

opts.

[alexveden]

I had a time on this week to finish swarm viewer report.

https://10.0.1.2:8889/notebooks/alphas/Alpha%20DSP%20BandPass%20-%20genetic%20optimizer%20swarm%20viewer.ipynb

I hope that this what are you expecting to get.

[nikolas-joyce]

I have played with the swarm view and i have some questions on what i am seeing... https://10.0.1.2:8889/notebooks/alphas/Alpha%20DSP%20BandPass%20-%20genetic%20optimizer%20swarm%20viewer-Copy2.ipynb

i am examining a swarm with the current values.

i get this outcome...

when i go to the Alpha DSPBandPass notebook to test and deploy the calibration here i get this outcome... https://10.0.1.2:8889/notebooks/alphas/Alpha%20DSP%20BandPass%20-%20genetic%20optimizer-Copy1.ipynb

is this the expected outcome ? Looks weaker than the swarm would suggest? Am i misunderstanding the selection mechanism?

[alexveden]

is this the expected outcome ?

This is possible, because alpha version is look-ahead bias free. I believe that is the bad sign, that is mean that the alpha tend to be curve fitting.

[nikolas-joyce]

But isn't the outcome worse than any of the individual swarm members?

[alexveden]

Also outcome depends on position management, and can be different comparing to raw equity outcome

Отправлено из Mail.Ru для Android четверг, 03 августа 2017г., 22:12 +04:00 от NikolasJoyce notifications@github.com :

But isn't the outcome worse than any of the individual swarm members? — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub , or mute the thread .