IBM Data and AI Ideas Portal for Customers

Shape the future of IBM!

We invite you to shape the future of IBM, including product roadmaps, by submitting ideas that matter to you the most. Here's how it works:

Post your ideas

Post ideas and requests to enhance a product or service. Take a look at ideas others have posted and upvote them if they matter to you,

  1. Post an idea

  2. Upvote ideas that matter most to you

  3. Get feedback from the IBM team to refine your idea

Help IBM prioritize your ideas and requests

The IBM team may need your help to refine the ideas so they may ask for more information or feedback. The product management team will then decide if they can begin working on your idea. If they can start during the next development cycle, they will put the idea on the priority list. Each team at IBM works on a different schedule, where some ideas can be implemented right away, others may be placed on a different schedule.

Receive notification on the decision

Some ideas can be implemented at IBM, while others may not fit within the development plans for the product. In either case, the team will let you know as soon as possible. In some cases, we may be able to find alternatives for ideas which cannot be implemented in a reasonable time.

Additional Information

To view our roadmaps:

Reminder: This is not the place to submit defects or support needs, please use normal support channel for these cases

IBM Employees:

The correct URL for entering your ideas is:

Status Under review
Workspace Cloud Pak for Data
Created by Guest
Created on Jun 16, 2020

Make Bayesian Optimization parallelize more

Currently the implementation of Bayesian Optimization in WMLA is completely sequential. This renders it essentially useless for models with long training times. It would be ideal to parallelize it more.
There are two angles to this: first, at its most basic level, it should be possible to specify starting the method by launching an initial number of trials at random or generated from one type of designed experiment (I believe the latter is how it currently works, it just doesn't execute them in parallel). The implementation of TPE already works in this way, so that it should not be much of a problem. Second, after this initial set of trials rather than proceeding in a fully sequential mode, one could run next not the top most promising point, but the k top most promising points (subject to a crowding constraint to prevent all next points to be picked from the same area -- I would be happy to work with you on the technical details of the selection heuristic if need be).