VFANN (Vector Fast Artificial Neural Netwoks)
The goal of this project is to develop a vector version of FANN which can take advantage of modern CPUs like those implementing SSE or similar technologies.
Another goal of the project will be to add utility apps/functions/scripts in order to make using FANN easier. A fann_trainer script or application will be in good shape at the end of the project and will be capable of training different kind of network with appropriate algorithms.
If projects for a FANN GUI are accepted I'll collaborate with the other student in order to make the utilities well integrated.
Vektörlerle YSA oluşturulacak sebebi ise modern işlemcilere daha uygun olması
Self-Organizing Maps and Growing Neural Gas (GSoC)
Competitive artificial neural networks have increasingly become popular for visualization and clustering of the large amounts of data existing in many scientific fields. They are also used for pattern recognition, image analysis, and many other applications. This project will add Self-Organizing Maps and a similar dynamic algorithm called Growing Neural Gas into the FANN library. The implementation will be coded so as to support extensions such as adding multiple learning rules and neighborhood functions, and include features such as calculation of quantization error and other metrics.
Self-Organizing Maps ve bir benzeri sayılabilecek Growing Neural Gas tipinde YSA lar da FANN kütüphanesine dahil edilecek.
Discrete-Time Recurrent Networks (GSoC)
Recurrent networks are an important feature currently missing from the Fast Artificial Neural Network (FANN) library. Not only can they be used to model new problems, but they also better mimic the connectivity of biological neurons. For this project, the FANN library will be extended modularly to add support for discrete-time recurrent networks. The analogues of the feedforward training algorithms which FANN already supports -- recurrent backpropagation and recurrent cascade-correlation -- will be implemented. Unique to recurrent networks, Long Short-Term Memory will also be implemented to provide a method which can learn to store information over long time periods faster than recurrent backpropagation. This solution will include documentation, test cases, and a tutorial.
yine değişik tipte bir YSA FANN kütüphanesine dahil edilecek.
Conjugate gradient training methods support (GSoC)
Conjugate gradient algorithms are popular training methods in artificial neural networks. Due to speed and average memory requirements they are often good choice especially for large networks. This project consists of implementation of three such methods: Fletcher-Reeves, Polak-Ribiere and Powell-Beale.
Conjugate gradient öğrenme algoritmalarından Fletcher-Reeves, Polak-Ribiere ve Powell-Beale ekleniyor
OpenOffice.org Spreadsheet Plugin (GSoC)
Artificial neural networks have proved to be extremely helpful tool for various kind of tasks, from basic pattern recognition to data processing. As there are many tools for working with neural networks, most of them is either very expensive or requires programming skills to write appropriate programs that create and train networks. As FANN proved to be the de facto standard library in the FLOSS world for neural computation, its usage is uncomfortable and counterintuitive for many people because of the lack of a GUI. (Commercial Mathematica bindings are not considered because of the price of the product) Creating OpenOffice FANN plugin may help to widen even more the popularity of the library and provide a means of performing neural calculations to people without necessary programming background.
Open Office tablolama programı ile FANN kütüphanesinin kullanabilmek için eklenti
Native GUI (GSoC)
Using QT became possible to have also a real cross-platform native GUI, especially with QT4 (windows, linux, macOs). Goal: My objective is to create a user-friendly cross-platform native GUI using Qt4 libraries.
Qt kütüphanesi ile yazılması planlanan FANN Kullanıcı arabirimi benim yazdığım FannTool benzeri bir program
