 |
| Home Alewife AlphaServer SC AP 3000 AV 25000 The Challenge CM-2 CM-5 Dash E10000 Exemplar V21000 Exemplar V2600 Flash Gamma II Plus Hector ILLIAC IV iPSC KSR-2 NCUBE/ten NUMAchine Paragon RS/6000 SP SR8000 SX-5 T3E VPP5000 |
| SNNS is a software simulator for neural networks on Unix workstations developed at the Institute for Parallel and Distributed High Performance Systems (IPVR) at Univestität Stuttgart. The SNNS simulator consists of two main components; simultor kernel written in C and graphical user interface under X11R4 or X11R5. The simulator kernel operates on the internal network data structures of the neural nets and performs all operations of learning and recall. It can also be used without the other parts as a C program embedded in custom applications. It supports arbitrary network topologies and can be extended with user defined activation functions and learning procedures written as C functions and linked to the simulator kernel. Currently the following learning procedures are distributed: - backpropagation (online, batch, momentum term, flat spot elim.) - counterpropagation - quickprop - backpercolation 1 -generalized radial basis functions (RBF) Version SNNSv2.2 includes ART1, ARTMAP and ART2 networks, RPROP, cascade correlation and recurrent cascade correlation. Time delay networks, Elman networks and some other network paradigms are completed but scheduled for a later release. The graphical user interface XGUI (X Graphical User Interface), built on top of the kernel, gives a 2D and a 3D graphical representation of the neural networks and controls the kernel during the simulation run. In addition, the 2D user interface has an integrated network editor which can be used to directly create, manipulate and visualize neural nets in various ways. A massively parallel version of SNNS for the MasPar MP-1 is completed but not publicly available. With 384 M CPS (connections per second) and 129 M CUPS (backpropagation) on our MasPar MP-1216 with 16,384 processors it is currently among the fastest neural network implementations on general purpose parallel computers. Parallel SNNS versions for the Intel Paragon, Alliant FX/2800, Kendall Square KSR-1 and others are planned. SNNSv2.1 is available free of charge for research purposes under a GNU-style copyright agreement. See the license agreement in the user manual and in the file Readme.license of the distribution for details. SNNS is (C) Copyright Univestität Stuttgart, IPVR. Note that SNNS has not been tested extensively in different computer environments and is a research tool with frequent substantial changes. The SNNS distribution includes full source code, installation procedures for supported machine architectures and some simple examples of trained networks. The current distribution supports color or B/W Sun 4 (Sparc) systems and DecStations under the original X11R4 or X11R5 (Athena widget set, twm, MIT fonts) or Sun OpenWindows 3.0, IBM RS/6000 systems with AIX and HP 9000/7xx systems with HP/UX 8.0.7 and HP VUE and 386 or 486 IBM-PC clones under Unix (e.g. Linux) and X11R4 or X11R5. SNNS has currently more than 300 users worldwide, approx. one third of them in Germany, in other European countries, and in the U.S., with a few users even in Australia and Japan. The feedback from users id mostly very positive. A Workshop "Simulation Neuronaler Netze mit SNNS" at the Universität Stuttgart, IPVR, Sept. 1992, was very successful. The workshop proceedings (in German) can be obtained from Universität Stuttgart, Fakultät Informatik, Dr. A. Zell, as Bericht Nr. 10/92. |