WE BELIEVE IN AUTOMATING EVERYTHING, INCLUDING AUTOMATION
ALGiT stands for the –” ALGorithm information Technologies”. Algit is a research and development based enterprise in a less known, but extremely important new field of Evolutionary Artificial Intelligence (EAI), which is our tool and our product at the same time. We believe that virtually everything can be optimized or improved, and we are constantly proving that, by applying the EAI on all kinds of optimizations and even on EAI algorithms themselves.
By the same approach we are investigating paths to programming automation. It is a way of software programming by bringing algorithms and software chunks inside our software simulator and enhancer. Selecting the best among many randomly mutated/tested, through billion of dependent generations, gets us a new and unique, previously unknown results. This is the way a machine/program can develop software for you, and not merely run the programs made by humans.
We have developed a system for evaluating algorithms. Algorithms represent the back bone of every computer program. Our product (Critticall) simulates the process of natural selection. The development of an algorithm is completely automated, due to the EAI. The fundamentals of artificial intelligence are clearly explained in generally accessible in the following documents ( Levin Search and Solomonoff induction). We approximate those methods by employing the digital evolutionary approach.
Crtticall is the first and so far the only working program for generating and evaluating a general algorithm and it has reached the full level of automation. For more information please refer to the technological pages of Algit. Some practical examples of this program are: improvement of the existing sorting algorithm written in C++ programming language and automatic generation source codes for 2nd and 3rd Kepler `s law.
Algorithms are not alone in this digital or simulated evolution process. For demonstration purposes, we have developed a program that shows the power of our concept, Pack´n´tile. The program optimally packs geometric shapes into other geometric shapes, using the power of the evolutionary algorithms. The results are obvious and understandable to everyone. The pages of reference where mathematicians challenge conventional method are Packomania or the Magdebourg University or Hugo Pfoertner. Pack´n´tile has already surpassed some of the world records and can reach or surpassed any record if that is theoretically possible.
On the basis of the Evolutionary Artificial Intelligence we have already developed the following products:
- ALGiT Time Table – is the most important among them, designed to calculate school schedules without any human assistance, considering a predetermined set of constraints.
- WoShi – employees schedulling
- Spector – deep data mining inside tables and texts
- Akritos – prediction of the next value considering what had already happened
- Pack´n´tile – optimal use of space by geometric figures and bodies
- OCR – Optical Character Recognition within texts and pictures
In the sampling process, the fundamental of any prediction or pattern analysis, our strength is the fact that we do not use any in advance chosen method but the evolutionary artificial intelligence finds the most appropriate algorithm, for each particular case. When we perform the pattern searching, which is prerequisite for the predicting and for the data analyzing, we do it by evolving the best fitted and the simplest (shortest in the Kolmogorov’s sense) algorithms which generates that particular data.
ACKNOWLEDGEMENTS AND SUCCESSES
Algit is also award winner for innovation at the Slovenian forum of innovation.
- ALGit TimeTable – Innovation ID I-195-07
- Evolutional optimization of algorithms – innovation ID I-437-08
Say, that we have the following observational data:
|Planet||Aphelion 000 km||Perihelion 000 km||Orbit time days|
More about it on LessWrong.
Automatic programming – Kepler’s law
The minimal, the maximal distance between a planet and the Sun (both in thousands of kilometres) and the number of (Earth) days for one revolution around the Sun. Above is only the empirical data and no binding algorithm among the three quantities. The celestial mechanics rules which go by the name of the Kepler’s laws. Can those rules be (re)invented by a computer program and how? After several hours the following C code has been evolved:
A part of this story begins with Johannes Kepler and his conjecture, proved only recently as the Hilbert’s 18th problem’s solution, how to optimally pack cannon balls in a presumably infinite space. It turned out, that the best way was in fact discovered by soldiers, even long before Kepler (Universiti of Pittsburgh).
If the space isn’t unlimited, what usually isn’t, a large set of newly opened problems spring. Many of them solved, even many more, unsolved to this day. Often, we have a reasonably good solution for which nobody knows if it is in fact the best or if it is not. Despite that solutions are often “the best known so far”, a huge academic and hobbyist community is out there in a search for ever better, more dense ways, to pack bodies and shapes inside some larger bodies and shapes. Books, papers and Internet sites are devoted for this quest of tide packing. Not only in 2D and 3D, but in higher dimensions as well.
Our method for finding the densest packing is the hard core evolutionism translated for the abstract virtual world of geometrical objects. For this kind of software the label Artificial Intelligence may be rightfully used, as it solves the tasks even humans sometimes can´t. At least that well. Not only individual humans, but groups of them, armed with computers and supercomputers and strong mathematical theories are no match for our humble Pack’n’tile, a Microsoft Windows based tool. It took decades to establish a particular solution which Pack’n’tile reproduces — or surpasses when possible — in a much shorter time. Pack’n’tile doesn’t relay on any knowledge at all, except that the shapes must go inside a smallest possible space and must not overlap.
We concede the importance of the past efforts, we only give you a new powerful tool. It will not prove you any theorem, except showing how more than thousand packings could be done. Unforeseen solutions may prompt you to investigate further, just get yourself an edge by having Pack’n’tile.