PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 27 Jun 2017 05:20 PM by  Deborah K. Sacrey
May 22, 2017 Workshop Follow-up - Mining Big Data
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages
Patrick Ng
Basic Member
Basic Member
Posts:151


--
26 Jun 2017 10:24 PM
    Recall in Deborah Sacrey's presentation, neural network was able to seek out features from 2-ms data, but not 4-ms, when 2-ms is resampled from 4-ms data. For some, that was puzzling.


    Q1: is such observation common, and if so, is there a simple explanation?


    Q2: what is the implication on Deep Learning and sampling rate?

    Thanks in advance.
    0
    Deborah K. Sacrey
    New Member
    New Member
    Posts:1


    --
    27 Jun 2017 05:20 PM
    Patrick: when one up-samples the data ONE time (more than that introduces artifacts), you are essentially doubling the statistical information for the computer to analyze. What I have found is that there are buried patterns in these additional statistics which were not present in the larger sample rate. This is very common in the work I am doing. By using sample statistics instead of wavelet statistics, one is able to get below seismic resolution to see stratigraphic units not obvious in the regular PSTM data. The implication to this is resolving thin reservoirs and subtle features in the subsurface not previously detected. Hope this helps. Best, Deborah
    0
    You are not authorized to post a reply.