Comments for Commercial Intelligence http://haleyai.com/wordpress systems that know and understand and think and learn Fri, 20 Jan 2012 02:37:12 +0000 http://wordpress.org/?v=2.9.1 hourly 1 Comment on Pursuing a decision tree down a rat hole by Peter Lin http://haleyai.com/wordpress/2012/01/19/decision-trees-as-rat-holes/comment-page-1/#comment-299 Peter Lin Fri, 20 Jan 2012 02:37:12 +0000 http://haleyai.com/wordpress/?p=369#comment-299 I totally agree a huge decision tree isn't a great model for effective visualization. Having used Blaze and other decision tree tools, once a tree is larger than 2 x 2 page, business users have a really hard time understanding the bigger picture. A tree with thousands of nodes will be difficult for programmers to understand, let alone business users or business analysts. I do think there's room for better visualization, but a huge decision tree definitely isn't it from first hand experience. I totally agree a huge decision tree isn’t a great model for effective visualization. Having used Blaze and other decision tree tools, once a tree is larger than 2 x 2 page, business users have a really hard time understanding the bigger picture. A tree with thousands of nodes will be difficult for programmers to understand, let alone business users or business analysts. I do think there’s room for better visualization, but a huge decision tree definitely isn’t it from first hand experience.

]]> Comment on Goals and backward chaining using the Rete Algorithm by paul@haleyAI.com http://haleyai.com/wordpress/2008/03/11/goals-and-backward-chaining-using-the-rete-algorithm/comment-page-1/#comment-255 paul@haleyAI.com Sun, 17 Jul 2011 14:08:58 +0000 http://haleyai.com/wordpress/2008/03/11/goals-and-backward-chaining-using-the-rete-algorithm/#comment-255 This sounds interesting and helpful, Mark. You may be interested in considering the use of tabling in prolog (called SLG resolution) which can be used to achieve Rete like benefits, too. It's good that you have unification (apparently when answers are received by Rete). It sounds like you have a simple recursive prolog under the hood, but in my experience they are very easy to "blow up"/loose control of when used to any significant degree by less than experienced logic programmers. Recent changes to XSB, which has SLG resolution, too, address quite a few of these practical issues. I'd be interested in your experience as this rolls out. Good luck with the rest of "opportunistic"! This sounds interesting and helpful, Mark. You may be interested in considering the use of tabling in prolog (called SLG resolution) which can be used to achieve Rete like benefits, too. It’s good that you have unification (apparently when answers are received by Rete). It sounds like you have a simple recursive prolog under the hood, but in my experience they are very easy to “blow up”/loose control of when used to any significant degree by less than experienced logic programmers. Recent changes to XSB, which has SLG resolution, too, address quite a few of these practical issues. I’d be interested in your experience as this rolls out. Good luck with the rest of “opportunistic”!

]]> Comment on Goals and backward chaining using the Rete Algorithm by Mark Proctor http://haleyai.com/wordpress/2008/03/11/goals-and-backward-chaining-using-the-rete-algorithm/comment-page-1/#comment-254 Mark Proctor Sun, 17 Jul 2011 06:45:34 +0000 http://haleyai.com/wordpress/2008/03/11/goals-and-backward-chaining-using-the-rete-algorithm/#comment-254 The other subtle different is that all query results are local to the caller. So a rule or a query calling another query does not insert any results into the working memory, so it's very "prolog like". We'll be looking to OBC to be able to have BC insert or update WM instances. The other subtle different is that all query results are local to the caller. So a rule or a query calling another query does not insert any results into the working memory, so it’s very “prolog like”. We’ll be looking to OBC to be able to have BC insert or update WM instances.

]]> Comment on Goals and backward chaining using the Rete Algorithm by Mark Proctor http://haleyai.com/wordpress/2008/03/11/goals-and-backward-chaining-using-the-rete-algorithm/comment-page-1/#comment-253 Mark Proctor Sun, 17 Jul 2011 05:50:27 +0000 http://haleyai.com/wordpress/2008/03/11/goals-and-backward-chaining-using-the-rete-algorithm/#comment-253 We finally got backward chaining (BC) working in Drools, although opportunistic BC is still to do. The route we've taken is to introduce a "hybrid" syntax for prolog like terms and OO slots, based on RuleML POSL. The implementation is like prolog derivation queries and supports full variable unification at the point of invocation. It's stack based, so does not have a problem with recursion and blowing te java method stack up. Unlike prolog as you advocate in your article our query can be left "open" so you can receive reactive updates to the query view when the underlying data changes. http://blog.athico.com/2011/04/backward-chaining-emerges-in-drools.html http://blog.athico.com/2011/06/truth-maintenance-over-directed-graphs.html Still plenty more to do, but it's a good start :) We finally got backward chaining (BC) working in Drools, although opportunistic BC is still to do. The route we’ve taken is to introduce a “hybrid” syntax for prolog like terms and OO slots, based on RuleML POSL. The implementation is like prolog derivation queries and supports full variable unification at the point of invocation. It’s stack based, so does not have a problem with recursion and blowing te java method stack up. Unlike prolog as you advocate in your article our query can be left “open” so you can receive reactive updates to the query view when the underlying data changes.

http://blog.athico.com/2011/04/backward-chaining-emerges-in-drools.html
http://blog.athico.com/2011/06/truth-maintenance-over-directed-graphs.html

Still plenty more to do, but it’s a good start :)

]]> Comment on Rules vs. applications of knowledge by James Taylor http://haleyai.com/wordpress/2011/04/22/rules-vs-applications-of-knowledge/comment-page-1/#comment-243 James Taylor Tue, 26 Apr 2011 05:06:37 +0000 http://haleyai.com/wordpress/?p=334#comment-243 Paul Thanks for the kind words and the link. I do have some other videos on my channel - http://www.youtube.com/user/decisionmanagement - but for now these are somewhat random and driven by particular events/clients who have recorded me. I plan to develop some more coherent video assets but it's not done yet! One of the keys I have found in working with clients is to Begin with the Decision in Mind. Only by understanding the decision you are trying to manage or improve, by seeing clearly its characteristics and understanding how it can be measured and monitored can you take useful action. This action could be rules-centric, analytical, semantic or some combination of a number of techniques but the focus on the decision will pull the project through in a way that a focus on the technologies will not. James Paul
Thanks for the kind words and the link. I do have some other videos on my channel – http://www.youtube.com/user/decisionmanagement – but for now these are somewhat random and driven by particular events/clients who have recorded me. I plan to develop some more coherent video assets but it’s not done yet!
One of the keys I have found in working with clients is to Begin with the Decision in Mind. Only by understanding the decision you are trying to manage or improve, by seeing clearly its characteristics and understanding how it can be measured and monitored can you take useful action. This action could be rules-centric, analytical, semantic or some combination of a number of techniques but the focus on the decision will pull the project through in a way that a focus on the technologies will not.
James

]]> Comment on What is has always been going to be by Peter Lin http://haleyai.com/wordpress/2011/01/10/what-is-has-always-been-going-to-be/comment-page-1/#comment-191 Peter Lin Mon, 10 Jan 2011 17:13:18 +0000 http://haleyai.com/wordpress/?p=311#comment-191 I love the title of the post. It's a great tongue twister. Interesting food for thought. One area of temporal logic that I've been pondering on the last few years is temporal versioning and temporal distance. I'm interested to hear how you address the expression of intervals of time. I love the title of the post. It’s a great tongue twister. Interesting food for thought. One area of temporal logic that I’ve been pondering on the last few years is temporal versioning and temporal distance. I’m interested to hear how you address the expression of intervals of time.

]]> Comment on IBM Ilog JRules for business modeling and rule authoring by Peter Lin http://haleyai.com/wordpress/2010/02/21/ibm-ilog-jrules-for-business-modeling-and-rule-authoring/comment-page-1/#comment-136 Peter Lin Tue, 23 Feb 2010 23:56:17 +0000 http://haleyai.com/wordpress/?p=260#comment-136 I agree capturing knowledge is a good thing. I guess i'm just stuck on the details of how one does that in an easy and maintainable way. Without good tools, the knowledge base can become a burden. That would defeat the purpose of creating the knowledge base to begin with. I agree capturing knowledge is a good thing. I guess i’m just stuck on the details of how one does that in an easy and maintainable way. Without good tools, the knowledge base can become a burden. That would defeat the purpose of creating the knowledge base to begin with.

]]> Comment on IBM Ilog JRules for business modeling and rule authoring by paul@haleyAI.com http://haleyai.com/wordpress/2010/02/21/ibm-ilog-jrules-for-business-modeling-and-rule-authoring/comment-page-1/#comment-135 paul@haleyAI.com Tue, 23 Feb 2010 15:52:30 +0000 http://haleyai.com/wordpress/?p=260#comment-135 I agree that IBM is deeply innovative and look forward to its impact on Ilog over time, Dan. Peter, I am not "over the top" on SBVR. It is good to the extent that it emphasizes formal capture of knowledge (even though it remains un-executable and is linguistically impoverished). It's not about building tools. It's about understanding knowledge-level specifications above the implementation level. SBVR is closer to what business needs than production rule tools, even those with syntax that appears somewhat natural. But again, I would not limit my thinking to SBVR but consider its capabilities strategically. I agree that IBM is deeply innovative and look forward to its impact on Ilog over time, Dan.

Peter, I am not “over the top” on SBVR. It is good to the extent that it emphasizes formal capture of knowledge (even though it remains un-executable and is linguistically impoverished). It’s not about building tools. It’s about understanding knowledge-level specifications above the implementation level. SBVR is closer to what business needs than production rule tools, even those with syntax that appears somewhat natural. But again, I would not limit my thinking to SBVR but consider its capabilities strategically.

]]> Comment on IBM Ilog JRules for business modeling and rule authoring by Peter Lin http://haleyai.com/wordpress/2010/02/21/ibm-ilog-jrules-for-business-modeling-and-rule-authoring/comment-page-1/#comment-134 Peter Lin Tue, 23 Feb 2010 15:23:31 +0000 http://haleyai.com/wordpress/?p=260#comment-134 I'm still not convinced SBVR is the right specification for building knowledge base tools. IBM's Watson research facility works on a wide variety of areas like ABLE led by Joseph Bigus. My thought is it's better to build a knowledge base framework to a mature level, and then extract a standard out of it. Going the other way feels like it's doomed to fail. I’m still not convinced SBVR is the right specification for building knowledge base tools. IBM’s Watson research facility works on a wide variety of areas like ABLE led by Joseph Bigus. My thought is it’s better to build a knowledge base framework to a mature level, and then extract a standard out of it. Going the other way feels like it’s doomed to fail.

]]> Comment on IBM Ilog JRules for business modeling and rule authoring by Daniel Selman http://haleyai.com/wordpress/2010/02/21/ibm-ilog-jrules-for-business-modeling-and-rule-authoring/comment-page-1/#comment-133 Daniel Selman Tue, 23 Feb 2010 14:36:33 +0000 http://haleyai.com/wordpress/?p=260#comment-133 Hi Paul, IBM is a huge company with extraordinarily broad R&D interests. It would not surprise me at all if some of those interests translated into product innovations, whilst others will of course remain in the research domain for the foreseeable future. As a newcomer to IBM I am constantly amazed by the work being undertaken by IBM Research... In terms of SBVR: IBM has many products that would benefit from SBVR and IBM is a very active participant in the standardization effort. I cannot comment on product release plans, but IBM has a generally strong history of innovation and leadership in standards. Hi Paul, IBM is a huge company with extraordinarily broad R&D interests. It would not surprise me at all if some of those interests translated into product innovations, whilst others will of course remain in the research domain for the foreseeable future. As a newcomer to IBM I am constantly amazed by the work being undertaken by IBM Research…

In terms of SBVR: IBM has many products that would benefit from SBVR and IBM is a very active participant in the standardization effort. I cannot comment on product release plans, but IBM has a generally strong history of innovation and leadership in standards.

]]>