guedalia@bimacs.BITNET (David Guedalia) (11/14/90)
Hi, Has anyone seen or heard a definition for a Neural Network. Must a Nerual Net. have specific types of neurons? Are there any criterias for the hardware implementation of the NN. ? Would a Kohenen Feature Map be considered a NN? I am writting my thesis and need to prove that the ANN that I may build (one day) is 'officially' a NN. Thank you David
ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) (11/17/90)
In article <2491@bimacs.BITNET> guedalia@bimacs.BITNET (David Guedalia) writes: >Hi, > Has anyone seen or heard a definition for a Neural Network. Not really. In the literature, and talking with researchers, you will here "neural network" associated with almost any kind of real-valued massively parallel computations which are used to perform AI tasks. More formally, however, "neural network" is usually reserved for real brain circuitry (as opposed to "artificial neural network", reserved for massively parallel networks which people think up and program/build). The tem used for the field of working with artificial neural networks is refered to as "connectionism," which makes one envisioned interesting computation done by interconnected elements. Kohonen maps can fall into the "connectionist" label fairly well. The moral of the story here is that there are no hard-and-fast definitions of these terms, which is totally apt for the "fuzzy" and fault-tolerant nature of neural nets. -Thomas Edwards
pako@lut.fi (Pasi Koikkalainen) (11/18/90)
ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: >> Has anyone seen or heard a definition for a Neural Network. >Not really. In the literature, and talking with researchers, you will Well, I have been able to found a multitude of definitions for NN:s. These definitions, however, do not really specify what Artificical Neural Networks are (or are not). As a result almost anything can be presented as a neural network (and also has been presented). The most popular ANNs seemto have some common features: - Massively parallel - No external control mechanism (communication of PE:s only via message lines) - adaptation, learning rules. - paradigms are simle >Kohonen maps can fall into the "connectionist" label fairly well. One must remember that there are two definitions of what "connectionist" networks are. The definition here was that they are ANNs. The other definition is that they are higher level decision making networks, similar to semantic networks in AI, introduced by Feldman and in the PDP book; a subset of ANNs. Kohonen maps are lower level ANNs in that sense, but certainly this is a typical Artificial Neural Network too. Just wait few more weeks and I will publish my lic.thesis on this subject, you will be able to get your copy via postscript ftp file. -- * --------------------------------------------------------------------- * Pasi Koikkalainen * Lappeenranta University of Technology * P.O.Box 20, 53851 LPR, Finland
slehar@thalamus.bu.edu (Steve Lehar) (11/19/90)
Let me try my hand at this one... NEURAL NETWORKS =============== A neural network is a computational model that is inspired by observation of natural computational mechanisms. Natural architectures are fundamentally different from conventional architectures in that they tend to represent information in a distributed way, and to perform computation in a parallel analog manner that seems to be more fault tolerant and robust if the input information is somewhat ambiguous. Neural approaches work best in applications where traditional computation has performed poorly, usually because the data is ambiguous or the context has a large influence on the data, such as vision, speech and cognition. They generally perform poorly in realms where computers perform well, usually because the data is deterministic, clearly defined and well understood, such as word processors, spreadsheets, arithmetic computation. SPECIAL NOTE to the "there's no such thing as neural networks" folks: ============ Since most neural networks are implemented by computer simulation, (except the real ones, that is) there is of course some overlap between "neural" and "non-neural" models; very simple neural systems are similar to non-neural equations, and very large networks of conventional systems are sometimes similar to very small neural networks. The difference is really in the inspiration of the model- neural models tend towards simple computational units and lots of them, whereas conventional architectures have much more complicated units and less of them. The need for the separate term "neural" is that this approach has been so counterintuitive that it took a couple of decades of bashing our heads against certain insoluable problems to realize that the way it is done in the brain is very different from the way it is done in sequential computers. The reason conventional computers were initially so popular is that theirs is a more obvious, predictable and direct road to the solution (IF this AND that THEN theother) than the neural way (IF some of these AND some of those AND I'm in the right mood THEN perhaps a bit of theother). There are those who claim that so-and-so's neural model is nothing more than such-and-such a mathematical procedure, and therefore neural networks don't exist. To those I say, since the geometrical solution to a problem can also be solved algebraeically, therefore geometry (or algebra, take your pick) doesn't exist. Many mathematical problems can be solved in a variety of ways, which can be shown ultimately to be equivalent. It just happens that certain classes of problems are more easily solved with one technique than another, so it is important to match your mathematical tools to the nature of your problem to get the most results for the least effort. -- (O)((O))(((O)))((((O))))(((((O)))))(((((O)))))((((O))))(((O)))((O))(O) (O)((O))((( slehar@park.bu.edu )))((O))(O) (O)((O))((( Steve Lehar Boston University Boston MA )))((O))(O) (O)((O))((( (617) 424-7035 (H) (617) 353-6741 (W) )))((O))(O) (O)((O))(((O)))((((O))))(((((O)))))(((((O)))))((((O))))(((O)))((O))(O)
pluto@babymilo.ucsd.edu (Mark Plutowski) (11/20/90)
slehar@thalamus.bu.edu (Steve Lehar) writes: >Let me try my hand at this one... >NEURAL NETWORKS >=============== >A neural network is a computational model that is inspired by >observation of natural computational mechanisms. Natural >architectures are fundamentally different from conventional >architectures in that they tend to represent information in a >distributed way, and to perform computation in a parallel analog >manner that seems to be more fault tolerant and robust if the input >information is somewhat ambiguous. Neural approaches work best in >applications where traditional computation has performed poorly, >usually because the data is ambiguous or the context has a large >influence on the data, such as vision, speech and cognition. They >generally perform poorly in realms where computers perform well, >usually because the data is deterministic, clearly defined and well >understood, such as word processors, spreadsheets, arithmetic >computation. I believe this is a decent characterization, and provides enough historical background to motivate the definition, but let me try my hand at a definition: "Recall the usual definition of a ``network'' as a connected graph of computing elements (nodes) in which communication among nodes occurs along arcs connecting the nodes (connections.) A ``neural'' network, by analogy with the biological namesake, is obtained by placing restrictions on the type of information allowed to propagate along the connections, as well as upon the type of computation allowed within each node. Each node is allowed to compute a mapping from a set of inputs to a scalar output value. The instantaneous value of the activation propagated by a connection is allowed to be a scalar value." Note that the inputs to a node can be elements of any set, and so this definition does not preclude symbolic input information, so long as the transfer function of the node is well-defined over such a domain. Usually, though, the inputs are assumed to be a vector of a real space, since so many learning algorithms are derived analytically. However, many learning algorithms are happy with inputs being elements of a set, since after all the two-bit input set {0, 1} can just as easily be taken to be the set {off, on} by slight modification of the learning algorithm, viz, by the appropriate use of propositions defined over the input set. The definition of what a connection is allowed to propagate does not preclude time-varying information, nor does it preclude propagation of symbolic information encoded as a scalar value, say, such that the symbolic information can be appropriately decoded at the other end. Also, the definition of network does not preclude global broadcasting of information, since we have placed no constraints upon the connectivity. Nor does it preclude recurrent dynamics, since each node may retain a history of previous inputs internally, or be served by an external set of nodes whose purpose it is to retain historical information, say, by emulating a stack, queue, or even a time-averaged statistical summary. It is important to define a neural network as a computational model, as Steve Lehar did above, since a neural network is not defined by its implementation, but as an abstract way of characterizing either a particular computational set of hardware, software, or, according to some folks, wetware. Improvements to my definition are welcomed, it may not be sufficiently general to encapsulate everyone's idea of a neural network. But it certainly encompasses my own understanding of the term, given the nets I've seen reported on in the literature. That's my opinion, what's yours? (Qualification: some "connectionists" may maintain that this definition is strictly contained within their definition of a "connectionist architecture." I would not argue with that viewpoint.) -=-= M.E. Plutowski, pluto%cs@ucsd.edu UCSD, Computer Science and Engineering 0114 9500 Gilman Drive La Jolla, California 92093-0114