IBM Watson Data & AI - Structured Ideas

Welcome to the idea forum for IBM Watson Data & AI — our team welcomes any feedback, requests, and suggestions you have for improving our products! 

This forum allows us to connect your product improvement ideas with IBM product and engineering teams. 

Visual modeler interface for CPLEX in DSX

In our current DSX offering we enable programmatic access with Python API for CPLEX engine to solve optimization problems. Leveraging this capability requires a user to have deep technical and programming skills. Iterating across scenarios with modified constraints or objectives requires modifying code. To open up this capability to extended user base with less technical skills we need to offer guided interfaces to visually build optimization models. Very similar to the support we have for ML models. The current Dumbledore Beta intends to enable only a subset of the optimization problems (scheduling scenarios) with its natural language interface. And the dashboard visualization requires editing/coding in Javascript for the dashboards. To enable business users the end-to-end tasks for  Optimization models should be supported with visual, guided interfaces. This will extend reach of DSX to address optimization scenarios as well.

  • Guest
  • Nov 27 2017
  • Accepted
  • Attach files
  • Ferenc Katai commented
    November 27, 2017 14:47

    maybe this is related to this https://ibmwatsondataplatform.ideas.aha.io/ideas/IBMWDP-I-181

    as long as optimisation model creation is concerned, you can't spare some type of "coding" = converting the business problem to a set of math inequalities, goals = optimizastion model.

    so the simpler the language you can do that (i.e. closer to the actual math equations) the better. However, what I'd optimally see is a SPSS Modeler type of GUI where tons of NON-optimisation related functions are either drag&drop (e.g. DB connectors, some type of scripting logic, visualization of data/solution)) or simple clicking type of things (like project management), where the only thing the optimizastion modeler has to really "code" is the declarative model (prefereably in OPL, not in an API-based GPL).

    Dumbledore, in fact does not help/improve the modeling experience itself as it is providing python API for modeling only. This means 4 things

    A. users of other language investments (OPL, Java, etc) can't use it

    B. as a consequence Python has to be learned

    C. whatever it comes with python has to be accepted (e.g. performance = other packages has to be downloaded to improve it)

    D. whever a new thing arrives the investment has to be repreated in the new thing >>> e.g. GO was created in Google even if the Python creator worked for there for years. Different things (= languages) are good for something and not so good for something else = humans are advacing since we specialize not the other way around, that is, to search for the magical "hammer" who can screw-drive, drill, and hammer and the same time with the same efficiency... :-)

     

    so as a general idea in the light of my comment (and assuming the first line highlightling the connection to an obvioulsy similar/same request in spirit) above I vote for this as well...

  • Alain Chabrier commented
    December 18, 2017 09:45

    These 2 ideas are not related at all! They should be unmerged.

     

    One is about better support of OPL in DSX, as a transition for current OR experts using OPL to DSX.

    The other is about NL assistant being currently limited to a small subset of problems (scheduling problems) and this items suggests we extend this more domains and problems.

     

    You try to interprete this feedback to be a support of OPL but it is not. It is saying that NL is good as a non coding optimization experiences, but it needs to be extended to more domains.

  • Susara van den Heever commented
    December 19, 2017 17:18

    Agree with Alain that these are 2 separate ideas that should not be merged:  one idea around a visual guided NL interface, and another idea around an OPL IDE inside DSX. 

  • Alex Fleischer commented
    December 20, 2017 09:26

    I logged again https://ibmwatsondataplatform.ideas.aha.io/ideas/IBMWDP-I-218

    since it was deleted when I asked to remove the link with this wish