|
|
|
|
|
|
0 Replies and 402 Views
AI teaches itself - no manual labels required 402 0
Started by Patrick Ng
One billion public-facing Instagram photos were used to train an algorithm created by Facebook to learn to recognise images by itself. https://www.bbc.com/news/technology-56321296 Implication - it becomes more important than before when embarking on AI / deep learning projects, we shall ensure adequate QC and comprehend the underlying distribution. Understand what goes into the system and be prepared to flag surprises (e.g., what we cannot comfortably explain among ourselves). Scenario ...
|
|
|
|
0 |
402 |
08 Mar 2021 12:36 PM |
|
0 Replies and 408 Views
AI surprised Google developers - tacking like human 408 0
Started by Patrick Ng
https://www.bbc.com/future/article/20210222-how-googles-hot-air-balloon-surprised-its-creators Note - encoding (chaining of alpha-numeric) , e.g., 'Be5' to move a bishop, makes powerful feature engineering when training machine to play chess.
|
|
|
|
0 |
408 |
23 Feb 2021 10:12 PM |
|
0 Replies and 415 Views
How well do ML models generalize? 415 0
Started by Patrick Ng
One word - underspecification. https://arxiv.org/abs/2011.03395utm_campaign=The20Batch&utm_medium=email&_hsmi=111648800&_hsenc=p2ANqtz-_V9t7U6iMD9fTxiU11FZnzxNK-2AfPsVjtcMZ_iKqASG_JMnwJGBK_fPPBfrXjS3OIaY6VZgknnoRDKMOxKM6Gay6GGQ&utm_content=111648800&utm_source=hs_email Perspective - draw analogy between geophysical data processing (wiener-levinson,, homomorphic and minimum-entropy deconvolution) and ML models (logistic regression, neural network, LSTM, GAN), their ability to generalize d...
|
|
|
|
0 |
415 |
18 Feb 2021 01:39 PM |
|
0 Replies and 411 Views
Texas Freeze Out cannot stop Energy in Data 411 0
Started by Patrick Ng
Update - as we tally up the Texas and Oklahoma freeze out, and recognize challenges many of our members face in the aftermath, not to mention pipes bursts, water damages etc, the three societies will postpone the EiD 2021 kick off to a later date (tentatively slated for April). There is time for us to better prepare and engage a larger audience. Additional considerations will be given to broaden geographic reach, and highlight the contributions AI and machine learning make. Meanwhile, all ...
|
|
|
|
0 |
411 |
17 Feb 2021 09:17 PM |
|
0 Replies and 390 Views
Why Energy in Data 2021? 390 0
Started by Patrick Ng
As any emerging technologies, often we struggle through the initial period of cutting-edge and bleeding edge, before reaching wider adoption and wiser application. As in AVO (used to derisk prospects predrill) and depth imaging (credited for opening up the entire Gulf of Mexico deepwater subsalt play), “conventional wisdom” changed from convention to convention before we as an industry converged on a set of best practice based on play and geology. Anticipate the use of AI and machine lear...
|
|
|
|
0 |
390 |
12 Feb 2021 12:57 PM |
|