Every year at this time, we present the editorial lineup for the coming months. In addition to the topics listed below, we'll be offering up our regular monthly fare of object-oriented programming tools and techniques, solutions to Windows development problems, embedded systems programming projects, network programming, advanced algorithms, UNIX systems, nitty-gritty DOS, and more. Our basic approach remains unchanged: one programmer talking to other programmers, sharing ideas and techniques. And there'll always be a lot of source code. With all this in mind, here's what you'll being reading about in next year's Dr. Dobb's Journal:
If you have a particular article in mind on these or other topics, or if you'd like a copy of our author guidelines, send us a note (DDJ, 411 Borel Ave., San Mateo, CA 94402) or call (415-358-9500).
We've long believed that you should write every program to be network aware, even if you don't initially plan on running it on a network. To underscore the importance of this, in this issue we're launching "Networked Systems," a new monthly section focusing on network programming.
In the coming months, we'll be examining all facets of software development for networked environments, from LANs and WANs to open (and closed) systems. Again, if you have an article you'd like to contribute to this section or have problems you'd like to see us address, give us a call.
Each year we try to give you a bonus edition of DDJ that supplements your regular monthly issues. This year's special issue is entitled Dr. Dobb's C++ Sourcebook, and if you're a subscriber, you'll receive it at no extra charge with your December 1992 issue.
Included in this bonus supplement will be interviews with C++ language designers, discussions of why some programmers think C++ is better than Fortran for numeric programming, source code for some powerful class libraries, useful C++ utilities, and much more.
Neural nets are in the news again. For starters, Bellcore has announced it's developing a neural-net computer that processes 100,000 signals per second. The system eventually will be used for routing telephone calls, assigning radio frequencies for wireless telephones, and speech recognition.
Caere's soon-to-be-released Windows-based FaxMaster software uses neural nets to recognize and re-create characters lost or eroded during data transmission of low-resolution faxes.
Ricoh, the Japanese company known for printers and photocopiers, has announced a high-speed, hardware-only neural-network computer that will be used as an embedded controller for copiers. Ricoh, which is manufacturing its own 16-neuron chips, claims bragging rights to what it calls the first neural-net computer developed solely in hardware.
VeriFone has unveiled a check reader that processes up to 20,000 images per second at an accuracy rate of 99.6 percent. The VeriFone reader is built around the Synaptics I-1000 neural-net chip that's capable of 1 billion operations per second. The PC-based reader has the I-1000 on a daughterboard, plugging into VeriFone's Motorola 68HC11-based motherboard. VeriFone and Synaptics claim the check reader is the first commercial application of neural-network technology.
Hmm ... this might be news to Nestor Inc. and others who for years have been developing neural-net software for handwriting recognition, predictive modeling, and speech recognition. I'm sure VeriFone and Synaptics meant to say "the first commercial hardware application." If so, I'll then leave it to them and Ricoh to work out who's really on first with neural-net hardware.
Copyright © 1992, Dr. Dobb's JournalJanuary 32-bit Programming
February Cognitive Computing
March Data Structures and File Formats
April Algorithms
May Operating Environments
June ASM and Architectures
July Graphics Programming
August C/C++ Programming
September Numeric Programming
October Object-oriented Programming
November Debugging and Profiling
December Interoperability
Networked Systems
Dr. Dobb's C++ Sourcebook
Neural Net News, or Who's On First?