Welcome to the LAAL site
There is no part of contemporary human activity left untouched by both the need for and the desire to collect data. The consequences of such a fact are obvious - we are surrounded by and, in fact, we are immersed in an ocean of all kinds of data (a.k.a. measurements, images, patterns, sounds, web pages, tunes, x-rays and/or ct images, etc.) generated by various types of sensors, cameras, microphones, pieces of software and/or other devices. Humans can't handle usually ultra-large-scale data sets but, we can develop algorithms and methods for performing the heavy mining job of learning from data. This is what is tirelessly, carefully and efficiently performed at the LAAL but, while doing that, we are fondly thinking aboutFrancesco Petrarca's (Petrarch's) XIV century motto:
"I couldn't face making a merchandise of my mind."
In addition, the line in Proverbs 16-16 points at the very roots of our research motivations and scientific efforts:
"How much better it is to get wisdom than gold! And to get understanding is to be chosen above silver."
Once developed we do apply our precious models 'everywhere', notably in Bioinformatics, Medicine, Engineering, Science, E-commerce, World Wide Web (exploration, categorization, and mining), Time Series (Prediction), Images (compression) and, ..., well, elsewhere.
(The tiny icon above is the segment of Milić od Mačve's painting "Izučavanje Lepenskog pisma", standing for "Learning About, or Study of, Lepenski Vir's script", painted in 1987.)
Our newest algorithms and software (downloadable from Downloads here) include:
- OLLA for Kernelized (i.e., Nonlinear) Classification - Stochastic gradient based OnLine Learning Algorithm (OLLA) aimed at large datasets is a fast and accurate designing tool for SVMs and other alike NL classifiers in 'primal' domain
- GSVM - The ultra-fast code for training support vector machines based on geometric approaches to solving SVMs
- ALH - a MATLAB based software that implements the Adaptive Local Hyperplane algorithm for solving classification and regression tasks for small and medium data sets when a high accuracy is required
- LinearSVM - an extra-fast software for creating linear SVMs for ultra-large data sets
- ISDA - Iterative Single Data Algorithm, a software for implementing support vector machines (SVM) models
- SemiL - Semisupervised Learning software