...
- Introduction to Thoth's new work plan for the coming year
- Thoth goals and objectives discussions
- How many networks and what type of networks – Wireless, wireline, private, public, transport or access focus? Core? Customer Edge?
- Discussion to collect requirements for Network Intelligent Collaborative Innovation Project
- Resources to support project work – storage repository for the data
- Definition of data requirements
- Operator use case requirements
Minutes
I think one fundamental issue is the lack of real world data. Without the data form network operators networks we can teach our models only with simulated model what always result in biased result. Is there any activity in Thoth to collect datasets from operators? The answer is yes we are trying to get some of the telecoms to contribute real world data. There are several data sets available (Orange), but not enough. The data it turns out is highly proprietary and unique to each provider. There are two approaches, gathering real data and then testing algorithms against it, or creating test data. Probably need a combination of both. The Network Intelligent Collaborative Innovation Project hosted by China Mobile.
Discussion of who are the participants need to be involved in an Network Intelligence project – usual suspects – Telecom, researchers, vendors? Need to define a deep learning set of use cases. Base it on existing use cases. Beth Cohenwould recommend to focus on the network components, rather than the operations use cases. Will we use Kaggle to manage the modals and datasets or do we plan to use something else / build something new for this?