PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 23 Oct 2018 11:49 PM by  Patrick Ng
Salt Identification Challenge on Kaggle
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages
Patrick Ng
Basic Member
Basic Member
Posts:148


--
24 Jul 2018 07:33 PM
    Consider the rate of external innovation in machine learning algorithms exceeds that of internal organization-specific development.

    "To create the most accurate seismic images and 3D renderings, TGS is hoping Kaggle’s machine learning community will be able to build an algorithm that automatically and accurately identifies if a subsurface target is salt or not."

    Check out the Kaggle Salt Identification Challenge - click on this link
    https://www.kaggle.com/c/...tgs+competition+2018

    Join the fun.
    0
    Patrick Ng
    Basic Member
    Basic Member
    Posts:148


    --
    23 Oct 2018 11:49 PM
    Update - first place entry used residual networks (ResNet). For detail how they did it, see

    Link: https://www.kaggle.com/c/...nge/discussion/69291

    p.s. residual networks was introduced by Microsoft Research to train much "deeper" neural networks, when straight neural networks deemed not practically feasible (e.g., time and resource constraints).

    Reference: https://arxiv.org/pdf/1512.03385.pdf
    0
    You are not authorized to post a reply.