Deep Learning/Machine Learning Technical Interest Group (TIG)

Applying new analytics, neural networks, computational approaches using structured and unstructured data, and also training neural networks with supervised and unsupervised algorithms. Chaired by Patrick Ng and Andrew Munoz.
Deep Learning - Machine Learning TIG
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 19 Jul 2021 09:10 PM by  Patrick Ng
Azure, AWS and GCP Snapshot Q2 2021
 0 Replies
Sort:
Join to post a reply.
Author Messages
Patrick Ng
New Member
New Member
Posts:92


--
19 Jul 2021 09:10 PM
    Premise - usability of OSDU (open subsurface data universe).

    Scenario - want to access seismic and well log data, prep ML.

    Azure - Machine Learning Studio (classic) offers the best-in-class drag-and-drop machine learning builder. For those experienced in seismic data processing, the UI comes naturally. However, when it comes to OSDU, if accessible via OSDU connector, like Power BI, that will supercharge ML Studio to ML Nirvana. (Just add one OSDU box at the top to the rest of prep-ML workflow and done.)

    GCP - once we extract and are happy with the column features, AutoML Tables is a powerful ML companion (no need to spin up a Jupyter Notebook). Caution - the new and heavily pushed Vertex AI offers similar treat BUT not the easy user-control turn on-off selected columns as in AutoML Tables. On-off test is important for understanding the value of information and benefit (add or subtract column features, which implies cost of acquiring additional data or redirecting funds for something else). And quantify delta in the resulting ML scores. OSDU support noted https://community.opengroup.org/osd...s/R2.0/GCP

    AWS - while the latest OSDU API is available, the gap lies in the connector. Think HP printer driver. Had HP delivered the standard API, but each computer vendor had to provide the driver (for you and I to print), HP wouldn’t have sold as many cartridges. Wish list - AWS provides single-channel and multi-channel (prestack gathers) drivers in open-source repository (like Github). That will accelerate use and market adoption of OSDU. Not to mention unleashing the creative energy from AAPG-SEG-SPE members. Think how much that will boost the collective capability of delivering CCUS etc.

    Overall - across all three cloud platforms, ML is made easier and usability increases each period.

    Q: what is your perspective and experience on OSDU?
    0
    Join to post a reply.