Deep Learning, Machine Learning, and AI

[lrix]

Hi All,

Has anyone created any library or libraries and examples designed to exercise Eiffel with Deep Learning, Machine Learning, or AI in general?

It appears to me that Eiffel with SCOOP might be a prime candidate for this tasking. The DL, ML, and AI code appears to be largely back-end (i.e. non-web, non-mobile front-end). If Eiffel had such libraries and examples, tutorials, walk-through's, and such things, perhaps we could spur some activity in that community.

I have been drawn off for a small bit into other matters, but will continue working on updating, freshening, and adding to Eiffel documentation and training videos. The DL, ML, and AI seem an excellent arena for demonstrating Eiffel within that context. Perhaps providing those back-end processes with SCOOP'd code and Graphics Processors would be enough to spur some real interest.

What do you all think?

[Alexander Kogtenkov]

Hi Larry,

Indeed, using a strongly-typed language for AI and ML, where python is mostly used so far, might be a good idea. I do not know how the things are organized in the existing ML frameworks, and whether SCOOP will be necessary or not - that depends on applications and work organization. But using Eiffel as an orchestration mechanism for ML and as a language to provide strongly-typed API for AI sounds promising.

Alexander Kogtenkov

lrix <lr...@jinny.com>:

[lrix]

Thanks for the response!

Does anyone know if we have any AI experts who are also Eiffelists? Bertrand? Anyone at ETH?

I don't see why Eiffel ought not shine in this area!

[lrix]

Looks like Google has an excellent tutorial to get started with!

https://developers.google.com/machine-learning/crash-course

[Iordan Chahanov]

Hi,

We have some AI expertise, but nothing written in Eiffel yet, related to AI/ML.

Yesterday I wrote a code generator for clusters/classes/features generation, related to the initial basic structure of Neural Networks simulator.
(I love Eiffel, but no time for real SW development)

In general, for AI/ML projects, you will need GPU access and a lot of matrix operations.

There is an alternativee, based on data driven simulation.

Eiffel can be good at both.

Best regards,
Iordan Chahanov

--
Sent from my Android device with K-9 Mail. Please excuse my brevity.

[lrix]

An interesting find: Data sets upon which one can exercise ML.

https://archive.ics.uci.edu/ml/index.php

[Louis M]

Hi,

If that can be of any use; I had a student that create a little neural network library entirely in Eiffel. You can find it there: https://github.com/ZeLarpMaster/Neural-Network . Also, there is a little example where the neural network learn the XOR operator there: https://github.com/ZeLarpMaster/NeuralNetworkExample . Of course, it is minimalist, but it can be a starting point. Maybe he would be interested in working with the community to develop it further. I know that he did like AI a lot (without being an expert).

Good day,

Louis M

[Larry Rix]

Got this on external C compile output using Linux and gcc:

Preparing C compilation

Compiling C code in E1

/home/larry/.es/eiffel_user_files/18.01/precomp/spec/unix/EIFGENs/base-scoop-safe/W_code/preobj.o(Cobj8.o): In function `F236_6741':

(.text+0x4c3e1): warning: the use of `tempnam' is dangerous, better use `mkstemp'

C compilation completed