http://leon.bottou.org/ 2010-07-14T06:34:35-04:00 http://leon.bottou.org/ http://leon.bottou.org/lib/images/favicon.ico text/html 2006-12-13T18:15:39-04:00 http://leon.bottou.org/projects/crasm?rev=1166051739&do=diff Crasm is a portable assembler for several 8 bit processors, initially written in 1987 for wp>AmigaOS. It supports the 6800, 6801, 6803, 6502, and Z80 microprocessors. It comes with a number of small examples, including the 6801 firmware of my first homemade modem, and the famous 6502 Microchess program. text/html 2007-08-17T12:01:00-04:00 http://leon.bottou.org/projects/djvulibre?rev=1187366460&do=diff DjVu is a web-centric format and software platform for distributing documents and images. It can advantageously replace PDF, PS, TIFF, JPEG, and GIF for distributing scanned documents, digital documents, or high-resolution pictures. DjVu content downloads faster, displays and renders faster, looks nicer on a screen, and consume less client resources than competing formats. DjVu images display instantly and can be smoothly zoomed and panned with no lengthy re-rendering. DjVu is used by hundreds o… text/html 2007-12-18T10:50:24-04:00 http://leon.bottou.org/projects/kxen?rev=1197993024&do=diff KXEN's mission is to embed advanced machine learning into existing enterprise applications. KXEN suite of analytical tools is used by leading companies around the world. Their main strenght is speed, ease of use, and scalability. I am member of the KXEN Scientific Board. text/html 2009-07-30T11:53:11-04:00 http://leon.bottou.org/projects/lasvm?rev=1248969191&do=diff 1. Introduction LASVM is an approximate SVM solver that uses online approximation. It reaches accuracies similar to that of a real SVM after performing a single sequential pass through the training examples. Further benefits can be achieved using selective sampling techniques to choose which example should be considered next. text/html 2006-12-13T17:14:06-04:00 http://leon.bottou.org/projects/lush?rev=1166048046&do=diff Lush (<http://lush.sourceforge.net>) is an object-oriented programming language designed for researchers, experimenters, and engineers interested in large-scale numerical and graphic applications. It is designed to be used in situations where one would want to combine the flexibility of a high-level, weakly-typed interpreted language, with the efficiency of a strongly-typed, natively-compiled language, and with the easy integration of code written in C, C++, or other languages. text/html 2010-01-07T11:45:35-04:00 http://leon.bottou.org/projects/minilisp?rev=1262882735&do=diff Many years working with Lush have taken their toll. There are many things I really dislike doing in C or C++ when I know how easy they would be in Lisp. The essential lisp data structure is called the S-Expression. S-Expressions can represent lots of complicated things using a simple printable format. These are the main reason why many people find XML appealing. But S-Expression are much simpler. text/html 2010-06-01T14:47:58-04:00 http://leon.bottou.org/projects/neuristique?rev=1275418078&do=diff Neuristique was founded in 1988 by a dozen friends with big dreams. The mission statement was very long sentence that mentions the application of artificial neural networks, the development of artificial brains, and the exploration of space. We were very young and inexperienced. text/html 2006-12-14T10:25:20-04:00 http://leon.bottou.org/projects/objprelink?rev=1166109920&do=diff Objprelink is no longer useful. Objprelink is a code transformation tool that optimizes the runtime linking times of C++ shared libraries. Although objprelink was little more than a proof of concept, many people used it to speedup KDE2 back in 2001-2002. This is no longer useful since runtime linkers now contain optimizations similar in nature to those demonstrated by objprelink. Anyway, objprelink only works with ancient versions of the GCC compiler. text/html 2010-02-22T10:31:37-04:00 http://leon.bottou.org/projects/sgd?rev=1266852697&do=diff Stochastic Gradient Descent (SGD) has been historically associated with back-propagation algorithms in multilayer neural networks. These nonlinear nonconvex problems can be very difficult. Therefore it is useful to see how Stochastic Gradient Descent performs on simple linear and convex problems such as linear Support Vector Machines (SVMs) or Conditional Random Fields (CRFs). text/html 2010-01-07T12:09:37-04:00 http://leon.bottou.org/projects/svqp?rev=1262884177&do=diff Early Support Vector Machine Solvers Support Vector Machines (SVMs) were discovered in three steps: In the sixties, Vapnik and Learner propose the Optimal Hyperplane classifier. In 1991, Boser, Guyon and Vapnik add the Kernel Trick. In 1994, Cortes and Vapnik finalize the modern Soft Margin formulation of SVMs. text/html 2006-04-04T20:37:49-04:00 http://leon.bottou.org/projects/tl3?rev=1144197469&do=diff TL3 is a compact lisp interpretor that is an ancestor of Lush. Its copyright belongs to both me and Neuristique. It was initially used in the SN Learning System Development Tool. It was eventually agreed to release TL3 under the GNU General Public License. The lisp language is pretty clean with a full online documentation and a graphic toolkit. There is also a well documented extension mechanism for adding new primitives.