|
Post by sigurdur on Oct 24, 2014 0:21:41 GMT
Going to b a record year for the big 2 in USA. Corn/beans. Test weight is light on corn tho. It was a bit too cool in the corn belt. Talked to a friend from Iowa yesterday. Can't be true - it was the hottest year evah in particular September was unprecedentedly warm, we have that on the authority of NASA/NOAA. You guys in the fly-over states have got it all wrong GISS will send you a print out from a computer in New York that will show you the temperatures were not cool. Well, well. They can show me all kinds of graphs. But. I. Live. Here
|
|
|
Post by nautonnier on Oct 24, 2014 1:13:10 GMT
Can't be true - it was the hottest year evah in particular September was unprecedentedly warm, we have that on the authority of NASA/NOAA. You guys in the fly-over states have got it all wrong GISS will send you a print out from a computer in New York that will show you the temperatures were not cool. Well, well. They can show me all kinds of graphs. But. I. Live. Here That means you are using the low grade information from observation in the field; NASA and NOAA use the far superior computer models!
|
|
|
Post by sigurdur on Oct 24, 2014 1:17:31 GMT
Well, well. They can show me all kinds of graphs. But. I. Live. Here That means you are using the low grade information from observation in the field; NASA and NOAA use the far superior computer models! I admit, my thermometer doesn't do 1,000 of a degree. But my body does...
|
|
|
Post by slh1234 on Oct 24, 2014 17:08:38 GMT
I thought of this thread during the keynote this morning. I'm at a Machine Learning conference this week.
There was a contest for applications using machine learning on a specific platform. These had to be full, usable applications built on this particular platform, and part of the presentation at the awards was to talk about how long it took to build and deploy on the machine learning platform we are here for (a cloud platform - much more than I can explain in a paragraph).
The winning application was called "Dr. Pig." It was submitted by a team of 4 data scientists (2 of them actually interns) from China. It's a mobile phone app, but the model is hosted in the cloud. It's all about modeling for pig farmers in China - who tend to be small operations - to allow them to plan their pigs and production over the next six months. It takes into accounts such things as feed costs and trends, piglet capacity (which became a joke as the keynote speaker (a data scientists from India) wasn't sure what that meant), seasonally adjusted temperature projections, and a number of other factors. Projections and models are, of course, based on historical data.
I thought it was a very interesting concept. Machine Learning is a bit of a new thing for me, and it actually differs from what I understood from way back in my Computer Science days in University.
As to the accuracy: Since it is a new model, we'll see, but another concept presented by the chief researcher in the machine learning R&D I think points out a key difference between business and government scientists/researchers. He talked about all the places that modelling and machine learning was used in their research, and the goal was "minimizing the cost of failure" which he set up as 1) most hypotheses prove out to be false so we must 2) set up for fast falsification - falsify as quickly as possible so we can quickly move from things that do not work to things tht may. That actually reminded me of a quote from a customer about 8 years ago in the tech industry. He contrasted his time in government departments with his experience in private industry by saying "Government agencies have as many good ideas as private companies, but there is no force in place to kill the bad ideas in government agencies like there is in private industry." In my thinking the "minimise the cost of failure" is that motivating factor in the private sector.
|
|
|
Post by scpg02 on Oct 24, 2014 19:24:28 GMT
|
|
|
Post by slh1234 on Oct 25, 2014 13:25:17 GMT
AI is one area of machine learning, yes. But the guys at the conference I was at were more interested in applications such as intrusion detection, anomaly detection, failure rate projections, improving stability on complex platforms, price projections, capacity projections, etc.
|
|
|
Post by sigurdur on Oct 26, 2014 7:35:46 GMT
Sig, How did the season work out? Code: Wheat crop was excellent. Soybeans a bit too much early water. Potatoes average at best and edible beans slightly below average.
|
|
|
Post by sigurdur on Oct 30, 2014 18:33:39 GMT
|
|
|
Post by sigurdur on Nov 20, 2014 2:58:53 GMT
Thanks, has been tried in ND many years ago. All it did was create water problems. Not economical to say the least.
|
|
|
Post by sigurdur on Nov 21, 2014 3:55:13 GMT
|
|
|
Post by sigurdur on Dec 27, 2014 18:01:18 GMT
|
|
|
Post by sigurdur on Dec 27, 2014 18:03:13 GMT
|
|