Quote:

Original post by Timkin ...slightly OT... Quote: since pretty much anything can be represented by binary What is π (pi) as a binary string please? ;)

1100111110000000 (0111000001101001)?

(UTF-8)

Edit: Great, I take a couple of minutes looking through different encodings that have the "π" character in them and somebody else steals the joke... ;)

Quote:

Original post by Steadtler your tutorial ends up propagating misconceptions about ANN common among people who studied them for a little while. Chief among those misconceptions: 1- ANN work just like the brain!!! 2- ANN can potentially do anything!!! This end up doing more harm than good.

As I said, it was my personal opinion which I got from studying them. I'm not saying ANNs can do anything as they are right now. What I'm saying is that they have a LOT of potential and I believe that some day, we'll be able to design them so that they can work just like a human brain.

If you think about the words 'Artificial Neural Network', it basically just is another way of saying 'artificial brain' as we know for a fact that the brain is just a massively complex 'neural network'. All I'm saying is that I believe that someday, we'll be able to make an artificial brain that works the same way as a human brain. To get it to that point, ANN theory will probably have to change a lot, but in the end, if we ever make a true artificial brain, it will have to be implemented just like a real brain; using a massive network of neurons and thus, technically, it be an ANN.

I'm convinced that ANNs can potentially work just like real brains. I guarantee you that any new technology that allows us to create a true human-like artificial brain will fall under the category of ANN.

Tutorials aren't the place for your opinions, expectations, or beliefs. They're simple collation of facts and basic explanation about what's going on. You want to talk about what you think ANNs can or can't do, fine, but don't inject it into something meant to bootstrap beginners.

Quote:

Original post by Flashthinker Quote: Original post by Steadtler your tutorial ends up propagating misconceptions about ANN common among people who studied them for a little while. Chief among those misconceptions: 1- ANN work just like the brain!!! 2- ANN can potentially do anything!!! This end up doing more harm than good. As I said, it was my personal opinion which I got from studying them. I'm not saying ANNs can do anything as they are right now. What I'm saying is that they have a LOT of potential and I believe that some day, we'll be able to design them so that they can work just like a human brain. If you think about the words 'Artificial Neural Network', it basically just is another way of saying 'artificial brain' as we know for a fact that the brain is just a massively complex 'neural network'. All I'm saying is that I believe that someday, we'll be able to make an artificial brain that works the same way as a human brain. To get it to that point, ANN theory will probably have to change a lot, but in the end, if we ever make a true artificial brain, it will have to be implemented just like a real brain; using a massive network of neurons and thus, technically, it be an ANN. I'm convinced that ANNs can potentially work just like real brains. I guarantee you that any new technology that allows us to create a true human-like artificial brain will fall under the category of ANN.

Oh gosh. How many times must we read these words, which are among the realms of faith and whishful thinking. Apart from Promit's words, which I strongly agree with, forgive me if I give you a few words of advice:

-People like to give their methods impressive names.

-When left with nothing better, people like to justify their methods by making analogies with biology.

-Our understanding of the brain is feeble at best, and from what we do know, it doesnt work like the ANN you know of.

Im not saying ANN are bad or useless. Im saying justifying them or encouraging its use by making far-fetched analogies do more harm than good. Dont let your understanding of ANN or any technology be clouded by dreams and hope of potiential.

If you want to improve your tutorial, replace this rubish by situating ANN in the field of Machine Learning. Why would we need it? What are the advantages and disavantages of modeling a generic non-linear function by parallelizing linear components? What are the underlying assumptions? The answers to these questions your readers should learn before they learn how ANN actually work.

Quote:

Original post by Promit Tutorials aren't the place for your opinions, expectations, or beliefs. They're simple collation of facts and basic explanation about what's going on. You want to talk about what you think ANNs can or can't do, fine, but don't inject it into something meant to bootstrap beginners.

Fine... I removed the controversial sentence regarding what can be done with ANNs. I don't feel like removing the anatomical references because they are actually based on fact. Of course, biological neural networks differ greatly from ANNs, but I feel that it's important for beginners to know what the technology is based on.

Quote:

Original post by Flashthinker [...] what the technology is based on.

...an oversimplification of the way people thought the brain works in 1943.

Quote:

Original post by Promit Tutorials aren't the place for your opinions, expectations, or beliefs.

A tutorial without personal opinion is bland. In fact tutorials do have more right for personal opinion than textbooks. Funny thing is a personal opinion of person working in field, is better than a badly researched fact.
Tutorials are often done for introduction into subjects, so a personal opinions have place there. (This is also a reason whey there isn't only one working tutorial, multiple different angles are more helpful than a 100 copy of some standard work.)

Quote:

Original post by Steadtler Quote: Original post by Flashthinker [...] what the technology is based on. ...an oversimplification of the way people thought the brain works in 1943.

I disagree with this. ANNs are still a good model for neurons and modeling the brain -- the problem is that they are not detailed enough.

Quote:

Original post by Steadtler Quote: Original post by Timkin ...slightly OT... Quote: since pretty much anything can be represented by binary What is π (pi) as a binary string please? ;) π = 0000001111000000

Bah...no fair! You weren't supposed to catch on to that one! ;)

The other answer would be to write a function to compute pi and convert that to a binary string to input to a Turing machine (but you cant run it in finite time)! ;)

As for ANNs being a model of the brain...

The brain is a connectionist system... ANNs are connectionist systems... that's as close as one can reasonably say the two systems are related.

Quote:

Original post by Steadtler Quote: Original post by Flashthinker [...] what the technology is based on. ...an oversimplification of the way people thought the brain works in 1943.

Not THAT different
Call it what you want, it still classifies as an ANN.