I´m trying to modify the DeepLearn script to include external timeseries that could add new ways to explode this capability.
I have a timeseries downloaded from Quandl that it´s not daily but a sentiment value changing every few days.
The signal that I´m adding to the adviseLong function is just a simple return of the quandl series on that moment, scaled to 100.
The problem is that when I try to train the script I always get the error:

Error 046: MIN LookBack exceeded by 19 bars (99)
Error 014: MIN needs LookBack 99 > 80!
Error 014: MIN needs LookBack 99 > 80!

The LookBack is set to 100, and I´ve tried with higher values, but I keep getting that error.
Anyone knows what it could be?

The timeseries is retrieved from a csv downloaded in Quandl with:



and then used on the adviseLong function like:



The rest of the script is just the one coming with the default zorro installation.

Does anyone know how to add this external timeseries to the deeplearning algorithm so it can be trained and used?

Did you try using the dataFromQuandl function? That function seems to load the Quandl data only after the lookback period.

Check here
http://zorro-project.com/manual/en/data.htm

It's suppose to be only available with Zorro S, that it's not my case. I just wanted to test it properly before thinking on the purchase.
Any way to do the same with the dataFromCSV function?

If you check the link I sent, at the bottom you find the actual source code for dataFromQuandl. Just adopt to your need

A higher lookback period is not needed for data from Quandl, but probably for your normalize() call.

You won't have luck with your advise call either. You cannot train a neural net with a single feature.