@Dave Apache Samoa seemed like a good starting point as they’ve already implemented a set of algorithms as storm topologies, additionally regardless of whether that community is active , we could still take that code and iterate on it within heron as a good starting point. To that end I feel the biggest challenge with the machine learning initiative is figuring out the exact use cases and operationalizing ml topologies within heron.
Sent from my iPhone > On Jul 5, 2018, at 3:11 PM, Ning Wang <[email protected]> wrote: > > Hmm. Good question. Maybe not yet reaching out. > > > >> On Thu, Jul 5, 2018 at 11:49 AM, Dave Fisher <[email protected]> wrote: >> >> Hi - >> >> Has anyone reached out to the SAMOA podling? Or is their architecture >> inverted from that being proposed I’m not sure how well the SAMOA community >> is doing as they have had low activity since early this year. >> >> Regards, >> Dave >> >>> On Jun 29, 2018, at 11:01 PM, Ning Wang <[email protected]> wrote: >>> >>> Brief notes for the meeting on June 29: >>> >>> - We need to hook up heron with Apache samoa. Saikat to create new issues >>> in github. >>> - Create a slack channel: #machine-learning >>> - Let's add potential use cases in the design doc: >>> https://docs.google.com/document/d/1LrO7XRcMxJoMM83wjRd- >> Ov74VAaomA_mXOAhCStgGng/edit >>> >>> >>>> On Sat, Jun 23, 2018 at 3:44 PM, Ning Wang <[email protected]> wrote: >>>> >>>> Brief notes for the meeting on June 22th: >>>> >>>> - still studying the documents. >>>> --- https://mapr.com/blog/monitoring-real-time-uber-data-using- >>>> spark-machine-learning-streaming-and-kafka-api-part-2/ >>>> --- https://databricks.com/blog/2018/06/05/introducing-mlflow-an >>>> -open-source-machine-learning-platform.html >>>> --- https://eng.uber.com/michelangelo/ >>>> - stateful storage might need to be improved (data size) to support big >>>> state object which could be required by ML jobs. >>>> >> >>
