views:

1737

answers:

8

According to an answer from here, artificial neural networks are obsoleted by Support Vector Machines, Gaussian Processes, generative and descriptive models. What is your opinion?

+13  A: 

Neural networks are one method of "machine learning." Just because there are new technologies, doesn't mean the older ones are obsolete. There are quite a few applications for them, including risk assessment for financial businesses.

They're quite good at detecting patterns, so people still use them in applications that need that. I've found them useful for risk assessment myself, using them for determining whether a given customer would be a high risk for the company based on a large amount of previous training data. There may certainly be better methods for doing something like that, but I found a NN to be a perfectly acceptable solution, with good results.

Alex Fort
+6  A: 

Yes, they are. Neural networks' problems with getting stuck in local minima (i.e. finding a solution that's better than the one to the left, and better than the one to the right, and having no way of knowing that there's a far better solution a good distance off) are inherent to the methodology, and the effort required to even partially compensate for them is considerably greater than it takes to just use a methodology that works better.

chaos
I certainly hope my downvoter (and upvoters on Alex Fort's answer, for that matter) has written a neural network attempting to solve a non-trivial problem. Ever.
chaos
(-1) Your answer amounts to 'We no longer need stairs, we have elevators' Sometimes, the simpler, earlier methodology may fulfill your needs just fine.
That's true if the earlier methodology actually is simpler. Backpropagation neural networks being replaced by support vector machines is more like vacuum tubes being replaced by transistors than it is like your example.
chaos
Yeah, I agree that my analogy wasn't perfect. I was simply saying that you've written off every case. And you did it using the 'works better' phrase, which is pretty magical. It implies that another solution 'worked better' for you, which means in your case, you shouldn't use Neural Networks.
Neural networks are indeed very complicated and the outcome doesn't make up for it. Most pattern recognition people don't like NN, while SVM is very commonly used today. I would say chaos has a valid point.
ypnos
But, in other cases, its possible that Neural Networks might be perfect for someone. But then again, I'm not an expert. I just didn't like your absolutist answer, because you should always be open to other ideas, and you shouldn't be discouraging people from exploring their options.
I don't really agree, devinb. I'm fairly well convinced that if someone asks a question like "should I explore this option?", and I think the option is a dead-end waste of time relative to other, equally available options, I certainly should discourage them from exploring it. They're asking me to.
chaos
Very well put. However, in this case, with the enormously wild number of possible uses for Neural Networks, AI, and other options in the field, I think each option should be examined. Like Global Variables and GOTOs, even maligned concepts have their place.
Maybe. The overhead of exploring a given option for a particular problem is pretty bleedin' high, though.
chaos
@chaos @ypnos Even if SVM is better, it is probably easier for a doctor (who knows zilch about computational complexity) to relate to ANN, as it is based on something "naturally existing" (like nature-inspired search approaches). Popular or better? Quick or careful? Take your pick.
Amit Kumar
@chaos, I agree with your cautionary tone, I just disagree with your dismissive tone. I hope that clarifies my position.
Why does it matter how easily a doctor can relate to it? Is the doctor going to be writing one?
chaos
@devinb: Fair enough.
chaos
If doctor is not going to be writing one, he is going to be using one within another tool. Anyway, we probably want doctors to write code, or don't we?
Amit Kumar
If he's using it within a tool, he should not need to know or care what ML methodology the tool uses. And I want doctors to write code exactly as much as I want software engineers to perform surgery on me.
chaos
Using AI methods, doctors may probably be able to encode their "intelligence", and then even a software engineer can perform surgery :)
Amit Kumar
That was the big AI buzz of 1976: expert systems. Their promise has not exactly been realizewd to the extent you describe.
chaos
@chaos Fundamentally I agree with your stance (and even more after reading your answer). My argument (about doctor) probably does not make much sense. I am not much familiar with the maths behind neural networks. I understand we need more maths research, not mysteries.
Amit Kumar
I have implemented a large NN library with alot of functionality. NN's are not obsolete. SVM's are better for many problems though. If you are getting stuck on local minima then and you can't break through it then that is a limitation of your library. In my experience anyone who claims that one machine learning technique is "better" than others lacks and appreciation of the subtleties of the real world where there is usually no absolutes. Currently for a problem I am working on I am using a mix of SVMs and NNs.
Steve
+21  A: 

From this guy's paper here: http://www.inference.phy.cam.ac.uk/mackay/BayesGP.html ('Gaussian Processes - A Replacement for Supervised Neural Networks?') he states

"The most interesting problems, the task of feature discovery for example, are not ones which Gaussian processes will solve. But maybe multilayer perceptrons can't solve them either."

However, Kidney magazine suggests that

"In conclusion, although we understand that for special problems the ANN may still yield reasonable results, we argue that in general (from a theoretical perspective) and in particular (for the considered case study) support vector machine indeed outperform ANN."

Finally: www.cs.umu.se/education/examina/Rapporter/MichalAntkowiak.pdf

The Fig. 4.3 presents a comparison of the best results achieved by each method. It appears that much better results in classification were obtained using ANN than SVM. It seems also that ANNs are more resistant to insufficient data amount, because even for small set of Melanoma Maligna pictures results were satisfactory. That cannot be said about SVM, which had a problem with classification of above mentioned disease and mislead it with Melanocytic Nevus.

So, like pretty much everything in CS -- it's a matter of trade-offs and not is this the "best" but the "best for your particular problem"

Matt Rogish
+17  A: 

I think the phrase 'no longer fashionable' is more appropriate than 'obselete'. The fact is that the research community is just as susceptible to hype and fashion as any other community.

Neural networks were hyped a lot several years ago as one of the early AI technologies which was going to solve all the problems in the world. Neural networks have since experienced a backlash, partly because they are thought of as old technology that failed to live up to the hype, and partly because they are thought of as difficult to work with.

However, there is some very interesting newer research being done in 'deep learning' which, as far as I understand, is based on an efficient way of training neural networks with a lot of hidden layers. Some of the results being produced by this technique are very impressive.

Neural networks have been out of fashion for a while, but maybe it's time for a comeback?

StompChicken
I agree. The "publish or perish syndrome" and some subtleties in the social phenomena around peer review process accentuate these drawbacks. At least this is my outsider perception.
MaD70
A few years ago when going through the classwork for my masters, I got into the ANN class with several preconceptions and this high expectation and even some mystique around Neural Networks. Although the concept is "loosely" inspired by real neurons, I quickly realized that it's a fancy name for a statistical method.The mystique mask fell, which was good, because that opened my mind for other techniques such as VQ, SVM, HMM and so on.
Padu Merloti
+2  A: 

A good reference to NN and much more is Andrew Moore's tutorials "on many aspects of statistical data mining, including the foundations of probability, the foundations of statistical data analysis, and most of the classic machine learning and data mining algorithms"

Denis
+4  A: 

Strange conclusion which reminds me an historical precedent, the perceptron's case (the perceptron is a simple kind of artificial neural network):

... in 1969, Minsky co-authored with Seymour Papert, Perceptrons: An Introduction to Computational Geometry. In this work they attacked the limitations of the perceptron.
They showed that the perceptron could only solve linearly separable functions. Of particular interest was the fact that the perceptron still could not solve the XOR and NXOR functions. Likewise, Minsky and Papert stated that the style of research being done on the perceptron was doomed to failure because of these limitations. This was, of course, Minsky’s equally ill-timed remark. As a result, very little research was done in the area until about the 1980’s§. ...

§ Minsky and Papert are two pioneers of AI, so their opinion was much considered in that time. This was the classic symbolic vs subsymbolic debate in Artificial Intelligence.

In fact such limitation was easy to overcome simply by adding more than one layer of nodes (artificial neurons).

The moral of the story is that a technology can overcome its limitations even with a modest improvement. Case in point (with a not so modest improvment) Jürgen Schmidhuber's and colleagues recent work on Recurrent Neural Networks (RNN):

... Early RNNs of the 1990s could not learn to look far back into the past. Their problems were first rigorously analyzed on Schmidhuber's RNN long time lag project by his former PhD student Hochreiter (1991). A feedback network called "Long Short-Term Memory" (LSTM, Neural Comp., 1997) overcomes the fundamental problems of traditional RNNs, and efficiently learns to solve many previously unlearnable tasks involving: ...

MaD70
+6  A: 

Well shallow neural networks are certainly less popular since methods like SVMs can be as effective (or more) with less tinkering.

However, neural networks are still very much active and relevant, especially deep neural networks, known as Deep Belief Networks (DBN). DBNs come in 2 flavors: convolutional, and restricted Boltzmann machines (RBM). Convolutional networks are typically used for vision (and I know virtually nothing more about them). DBNs built from several layers of RBMs are great at learning high-level features of data in an unsupervised fashion, autoencoding, semantic hashing, and yes, classifying.

The trick is that DBNs are pretrained before using back-propagation, which is typically slow and kinda useless past 2 or 3 layers.

Two great sources:

Hinton Video Tutorial

Learning Deep Architectures for AI

Junier
wow.. these are new to me. Thank you for sharing these with us. How did you find out about DBN?
lmsasu
:) I just happened to stumble upon it when looking for some material on neural networks a few months back. More recently, I trying to employ DBNs in a project of mines. It was only in 2006 that RBM DBNs were really made possible, so it's still really new.
Junier
A: 

Relatively small neural networks of the type built to date might be considered (by some) to be unpromising and therefore obsolete. On the other hand, a neural network with around 100 billion nodes with about 100 trillion interconnections (ie, something on the scale of the human brain), might be surprisingly effective.

joe snyder