/* ---- Google Analytics Code Below */

Wednesday, August 15, 2018

Seeking Cognitive Infrastructures for Process ML

Have been reviewing work we did with machine learning as far back as 1998.  We initially installed and used inductive techniques to learn and adapt rules from data.  This was useful, but not for big data applications.   At the time we also chatted with SOAR which was working on the problem as well.  Our efforts ended.  But now I am reexamining the problem of general machine learning.

Convolutional Neural nets and their variants have been very successful for solving difficult, data greedy pattern problems.   But are not necessarily useful for dealing with complex business process  where the operations that need to be learned consist of many steps, and depend on complex context defined by data.

I also discovered that while most AI oriented companies we worked with then no long exist.  A  SOAR applying company still exists: (https://soartech.com). A quick scan of the work they mention seems to indicate much of what they are working on is for DOD and Government applications.

Wikipedia provides an overview of SOAR
The SOAR architecture still exists and is posted at a U of Mich site: https://soar.eecs.umich.edu/

Anyone have experience with SoarTech as a means of machine learning in process?  Willing to talk?  Collaborate in learning or use?  Contact me.   On Linkedin.   Will be diving deeper.

" ... At SoarTech, our focus is in the development of intelligent software that reasons like humans do, to automate complex tasks, simplify human-machine interactions, or model human behaviors. Our philosophy is three-fold: to be an augmentation to, not a replacement of, the human; to think “top-down, not bottom-up;” and to be transparent so that decisions and processing are communicated to the human and in human-like terms.  ... " 

( note that some of their efforts deal with drone swarms, also mentioned here.  I also note there is not much mention of 'learning' in the applications they mention)

No comments: