21 Oct Fuzzy Interconnectedness
Fuzzy and Interconnected Techniques
Section 5 suggests that the software of cognition is very fuzzy and able to operate efficiently even without having complete or totally accurate information. We said that we want to replicate that flexibility. We spoke in Section 7 about different fuzzy approaches for representing and processing information. These approaches include artificial neural systems and weighted conceptual schemata.
Variable certainty, confidence values, and weights all offer “fuzzy logic” alternatives to the rigid dichotomy of TRUE-FALSE logic. By applying weights to data elements, relations, and/or rules, we can overcome the brittleness of too much dichotomous logic and build more intuitive mechanisms for problem solving.
If we use neural networks, the structure for the information and the processing mechanism will be dictated by the implementation of the neural network. If, however, we turn to explicit KR schemata, we must match the KR scheme to the processing mechanism to provide fuzzy processing that can correctly and efficiently handle complex tasks.
In Section 7, we discussed one popular application domain that is intuitively well-suited to expert system techniques: the traveling salesman problem. Our choice of machine translation (MT) in this Section is an effort to show the usefulness of expert system techniques in a non-traditional domain. MT has traditionally been solved by “algorithmic” rather than fuzzy inference or neural techniques.
|Understanding Context Cross-Reference|
|Click on these Links to other posts and glossary/bibliography references|
|Prior Post||Next Post|
|Coefficient of Bureaucratic Drag||Methodology or Mythology|
|fuzzy logic confidence||Laguna 2004|
|dichotomous logic information||Pao 1989|
As stated in the section on design, the user interface is an important part of the overall design of any system. We must give users an intuitively easy way to get around. Hypertext, which was chosen as the primary navigation method for the world wide web on the internet, is one easy way to get around. For that reason, it was also chosen for the original Understanding Context content, which was created in Asymetrix Toolbook. You’ll note a fair number of links in the text of this post. This is not an accident, nor is it an accommodation. It is core to the premise of the blog that linking, based on words, is an extension of the associationist model described as early as Aristotle’s day.
Multimedia and Mobile computing have brought about a revolution in the AI community. Now, imaginal and two-dimensional geographic computing techniques can be integrated into expert systems of nearly any type. One of the best ways of integrating multimedia into expert systems is through internet style hypertext URLs or URIs. Hypertext is a multimedia technique that permits users to use a pointing device or keyboard to navigate from concept to concept, whether the concept is represented by text, sound, graphics, or even animated sequences of all of the above.
The Hypertext linked model is less and more than the relational model. It is less in that it does not include primary and foreign keys for conducting high-performance queries on large amounts of data. Hypertext implementations sometimes do not even provide a table construct. It is more in that any data element can become a “key” or link to any other data element in a non-linear structure.
Web style Hypertext applications are appealing to those of us with short attention spans. We ADHD types don’t like to spend any more time than is necessary on things that we know about already, or that we don’t feel a need to know. Hypertext links permit us to focus on a word or topic that appeals to our curiosity and travel that path until we reach the end of our attention span and choose to walk another path.
Down to Business
Modern spreadsheets are getting so smart that they can sometimes automatically figure out the best form of graph to use to show a series of numbers. Now that’s smart! Is it fuzzy? Just the opposite. Is it interconnected? The ability to associate text and numbers with pictures is powerfully fuzzy. When you hear “pie,” is your first association a delicious dessert or an expressive chart? Without knowing it, we often make our data very fuzzy. In the example here, “Sales in Millions” is ambiguous on several levels, but we implicitly understand what we’re looking at. Many people may not even need further explanation of the section labels, but they are collectively and individually very ambiguous.
In the corporate world, pie charts such as the Microsoft Excel * pie at right and bar graphs have become standards for representing certain types of data. Trends are often shown using X-Y graphs, and the direction and angle of a line’s ascent or descent can arouse strong emotional reactions in viewers and can even have influence across national borders. Wall Street watches London, Chicago, and Tokyo, and vice versa. “Blips” in New York can influence people in L.A., the EEC, Japan, and stock markets all over the world.
* Microsoft Excel is a registered trademark of Microsoft Corporation
The effort required to build intelligent machines that understand context is very significant. As we seek a strategy for enabling electronic devices to understand people’s intent from their words and the remainder of context, we are well served by considering not only how we establish these fuzzy interconnections between data, information and conceptual knowledge, but also considering how consumers and companies might benefit from such technology. The benefit discussion, I believe, is often fuzzy, but critical to the question of how to get the effort off the ground.
|Click below to look in each Understanding Context section|
|4||Perception and Cognition||5||Fuzzy Logic||6||Language and Dialog||7||Cybernetic Models|
|8||Apps and Processes||9||The End of Code||Glossary||Bibliography|