Fuzzy logic

Yes thanks. Know of any neural net calculators?

Pease's take on it, from 20 years ago.

First of 5 parts:

To the OP: what's changed since then in the fuzzy world?

Cheers

Phil Hobbs

July of 2010 was talking with this company regarding pattern recognition using IBM's neural network chip built into their cameras. It is my understanding that this chip has the ability to create it's own algorithm based upon reward/punishment learning sequences.

Their camera's are industrial, high quality

try to reach Matt Hori [say hello for me]

GEViCAM Inc.

Neural nets =/= fuzzy logic, AFAIK.

Cheers

Phil Hobbs

what is definition of that symbol combination?

"=/=" ??

=^= ?:-) ...Jim Thompson

-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | San Tan Valley, AZ 85142 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

ASCII art for "not equals".

Cheers

Phil Hobbs

You'd think, following many schematic entry conventions to make an "overbar", it'd be \=\ ...Jim Thompson

Or "!=" in many programming languages. "" in BASIC though. Or just use a modern encoding and put in "?" directly.

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

No, that's "defined as" (an equals with three lines). ;-)

But then there'd be '=#', the almost-correct '=!', and so on.

Tim

usiong Aioe hae to snip alot.

I snipped the person you replied who was asking about neural networks, oops

should have replied back up to that person not to you.

wow, I thought I saw you use that =/= somewhere and it meant, identically equal

now I can't remember what it was about.

use a lot of != ~= _=

but have never seen =/=

Just yanking... The actual math symbol is a single = sign with a / drawn thru it, thus =/= is the best we can do with ASCII. Although my Agent can replicate some of those symbols, they may or may not be readable on your end...

...Jim Thompson

"Identically equal", or "defined as" is usually "==".

It's quite common.

cally

As the one who mentioned neural nets, it is strictly true that ANN's are a different beast, but the distinctions are fuzzy, imho. The ANN mentor I had , Prof. David Rumelhart at Stanford, claimed that there was a great deal of overlap and implemented fuzzy in neural nets, he said. I'm an applied guy, so to me the net result of an analog signal coming out of a neural net (The Widrow-Hoff 3-layer with sigmoidal squashing function

- the work horse that most researchers use) is a "fuzzy" analog signal whic h takes on a value between .2 and .8 typically, as its best learning and fe edback range. So the output can be partially on if the net is not sure you have an "o" vs. a "c" for instance in ocr. The main benefit of ANN's is that they can be trained from data without OZi ng behind scenes with algorithms. A nice demo is the one that backs up a tr actor trailer from any start point to a loading dock. Compared to fuzzy logic, they offer easy machine learning, but the main cri ticism is you have limited knowledge of how they reached their result. In general, they are also good at reaching conclusions from many analog var iables, such as the 15 or so pathology tests in a cancer case.

I have a CHARACTER application, called Character Map Pro, that shows an array of potential characters, then click on whatever you want to find the proper alt-0ddd to produce that character, using 'default charset' everything is there, but 'not equals' sign

This browser even shows cyrilic(sp?) alphabet from Alex Bordodynov's emails *and* ?? characters from from unpronounceable Asian named people. and Korean alphabet, but no 'not equals'

Interesting, just saw a 'not equals' go by, a single character showing an equals sign with a line through it. I should have ctrl-c 'ed it for later use.

You brought to mind...As I was leaving undergraduate school, a math professor who knew I was interest in AI [he was withdrawn and extremely introverted type] suddenly talked to me. Invited me to his office whereupon he reached into lower right hand drawer of his desk and produced a print of paper he had written and handed it to me essentially saying, "When you get to Stanford, give this to Widrow. It's a proof of the convergence for his alogrithms. He's always ending his paper with, "it has been shown to converge" and he always leaves his papers that way. This should help him." So, at Stanford I sought out Widrow, whom I found to be quite affable, handed him the paper and told him the background of it. Whereupon, he instantly rummaged through it, grumbled something unintelligeble, and had his whole personality change to ascerbic as though I had ruined his whole day. Forgot all about that incident until your posting.

PS: Did you find anything out about the IBM 'learning' chip?

I suspect not a lot or we'd have heard more about it.

Indeed. Although NN can be made to implement fuzzy logic.

We had a fuzzy logic washing machine in Japan when it was all the rage. The thing would sometimes at random just rinse the clothes in cold water for no apparent reason. Their white goods were surprisingly fragile. It had to sit in its own sink to catch leaks and spillage!

Fuzzy did work incredibly well for breaking on superfast trains so that they always stopped exactly parallel to the platform queues though.

Windows has a program called Character map that lets you chose the font you're using, then displays every character in the font. You click on what you want, copy and paste it into whatever you're doing.