Machine Learning (ML) Usage Examples

[tdelbecque]

I have not solve my problem yet, and of course I will let you know if I find someting.

As for my environmment:

- I an running a thor cluster in Boca, provided and maintained by Lexis Nexis (we are a team working on it). So this is a rather orthodox cluster, though I do not know much about it;

- the common repository is hosted there

- I am using ECLIDE Version 6.10.4.101.3.8, but the problem is still there with the latest version;

- I got the latest ML package from github, and put it in the repository with AMT, but the same problem existed with a previous installation of the ML package.

nothing exotic, as you can see.

Bests.

[tlhumphrey2]

tdelbecque,

I don't know what your problem is. But, I'm trying to find someone who might know.

I'm fairly sure the problem isn't in the ml usage example you got from the ecl-samples repository because I ran yesterday and this morning without any problems.

Currently, I'm thinking something is wrong with the way the ML library is installed.

Do you work for a Reed-Elsevier company? If so, would you give me your work email address, so we can communicate faster?

Tim

[tdelbecque]

I am also absolutly sure that the example is fine, and that there is something specific to my settings.

I am indeed working on an Elsevier project with an Elsevier team, but as an external. My e-mail is thierry.delbecque (at) sodad.com.

Bests, Thierry.

[tlhumphrey2]

tdelbecque

Here are very high level instructions on how to install the ML library, or at least how I installed it.

First, get the ML library on your PC, the one where you run ECLIDE. For example, the path where I have the full ML library structure is: C:\Users\humphrtl\Documents\github\ecl-ml (By the way, on my PC, I’m running Microsoft Windows 7).

To setup ECLIDE so it will show the full ML library as part of the repository, go into “preferences” of the ECLIDE and click on the “compiler” tab. There you will find a text box titled “ECL Folders:”. Add the path to the ML library structure to that text box.

After doing that you will see the ML library structure in the repository.

And, in fact you will probably see two versions of it: 1) the one you have already installed and 2) the one that you just told ECLIDE about. You will have to remove the path of the currently installed ML library from “ECL Folders” text box.

Let me know if you need additional details.

[tdelbecque]

Thanks.

I am still investigating a bit on this funny problem: for example, if I rename Mul and remove Add in the Each attribute, then it works ... This makes me believe that there is something happening in the code generation and the name resolution. Should it be possible that it comes from our common repository ?

Bests

[tlhumphrey2]

tdelbecque,

Do you have more than one ML visable in the ECLIDE repository(or possibly a subfolder that appears more than once)? I'm thinking that might be your problem.

If this is the case, then you need to remove the path of one ML library from "ECL Folders:" text box in the "compiler" tab of preferences.

Tim

[tdelbecque]

I thought of that and I checked, and it seems not. Also I tried from too distinct machines, so too distincts ECLIDE installations, and got the same problem.

But I am going to check again, we never know ...

[john holt]

Theirry,
I think that the problem is an incompatibility between using the old shared repository on the pre-OSS world and the OSS.

To verify that this is indeed the problem, I would like you to clone your ECLIDE configuration, but in the clone make the Attribute server entry blank. You may need to selected Advanced to allow the entries to be changed.

You should of course have a copy of the attributes locally for the ECLIDE.

[tdelbecque]

Hello John,

Thank you. It seems that you are right indeed !

I did as you suggested, and the code compiles and executes when I set the attributes server to blank and use a local copy of ML.

So it seems that the issue comes from the pre-OSS version of the repository, finaly.

Thierry.

[jwilt]

Was there ever found another solution to this, besides a separate file-system code repository?
E.g., with IMPORTs, or compiler options, or #OPTION, ...?
Thanks!