Is this software of interest?

[WolfVector]

Andrew_st1:

>> “This resulting network is downloaded to the client machine upon completion and is not stored on the server. All the scanning takes place on the users PC using their own market data connection”

This is definitely a better business model. I assume that the output of this NN running on the user’s machine can be interfaced with a strategy running in TradeStation or NST? (E.g., a DLL type of interface supported by both). I assume that your NN just gives a buy/sell type of signal and that money management, position sizing, and protective stops would have to be set by the user.


>> “average learning time for the algorithm is about 4 hours . . . desktop machine this would be more like 16 hours . . .”

For this benchmark what is the size of the training set (number of bars) and how many inputs, I assume that the NN has one output (with something like 0 = ExitLong/Short; 1 = Long/ExitShort) (so I can compare it to NST).

--------

One final question, perhaps hard to answer without a standardized test, but have you found that your NN has significantly better performance than what NST (or other exiting commercial NNs) would provide using the same inputs in an out-sample-test?

 

[Andrew_st1]

Quote from WolfVector:

>>I assume that the output of this NN running on the user’s machine can be interfaced with a strategy running in TradeStation or NST? (E.g., a DLL type of interface supported by both). I assume that your NN just gives a buy/sell type of signal and that money management, position sizing, and protective stops would have to be set by the user.

Yes on both counts.

>>For this benchmark what is the size of the training set (number of bars) and how many inputs, I assume that the NN has one output (with something like 0 = ExitLong/Short; 1 = Long/ExitShort) (so I can compare it to NST).

Hmmm... I'm not too sure of those details at the moment. I'll get back to you on that when I get to ask our resident technical expert. I think something like 6 input, with 1 yrs data for ~20 stocks. (but the algorithm is going through a major optimisation process at the moment which may change all that.

>>One final question, perhaps hard to answer without a standardized test, but have you found that your NN has significantly better performance than what NST (or other exiting commercial NNs) would provide using the same inputs in an out-sample-test?

We haven't done those benchmarks yet. I shall get back to you with details when the optimisation process is completed and with a benchmark analysis.

Cheers,
Andy

 

[Pyrhus]

To Andtrew's


- When do you plan to market your new software ?

- What is it's name ?

- Have you a Website to read about it ?


Does your software use Fuzzy logic to code indicator inputs

to neural network , like Safir-X does ?


What are the mains difference of your product with Safir-X .


Thanks for your reply _

 

[damo484]

any new on pyrhus's questions.....

all this chatting has got me curious.

and... how would this compare tp Biocomp??

 

[bernie4bcw]

Hi Andy

How far are you with your project, have you gon commercial by now, if yes, what is your website, if no, let us know if the project is dead

Bernie