Nemoa is a python package for designing and building complex systems of stacked artificial neural networks. The key goal of this project is to provide an intuitive framework for designing data transformations which implement given requirements and also respect local and global restrictions of the underlying data structure. Common requirements for example are reconstruction of missing information, data classification, structural data analysis or data dimensionality reduction. For a current featurelist see the [changelog](https://github.com/fishroot/metapath/blob/master/changelog.md).
The designing process and the implementation of automatic intelligent systems like autonomous systems or data mining applications is very challenging and not guaranteed to result in success. Therefore the operational area of well designed intelligent systems is currently restricted to very few applications, mainly found in scientific, military or professional data mining environments. Also information about improvements is hardly shared between those groups and if indeed, than its very time-consuming and expensive to keep the own implementations up to date by single groups. The longterm goal of this project is to improve and also spread the use of automatic intelligent systems by providing a standard toolbox and also community specific frameworks that use this toolbox and focus on applications. This allows designing processes of intelligent systems, that are dominated by requirements and knowledge about the underlying data structures.
##### Licensing and Installation ##### Nemoa is available free for any use under the [GPLv3 license](https://www.gnu.org/licenses/gpl.html) and can be downloaded from [GitHub](https://github.com/fishroot/nemoa)
##### Contributing ##### If you are new to the field of [artificial neural networks](http://en.wikipedia.org/wiki/Artificial_neural_network) than let me give you a motivational example: Assume an automatic intelligent system for human disease detection which uses complex information like [gene expression data](http://en.wikipedia.org/wiki/Gene_expression) and a central storage and analysing system like nemoa, which continously optimizes and analyses large instances of artificial neural networks to detect and forecast human diseases from measurement of single patients or even whole communities. This is technically possible and not even very complicated! Of course the main tasks in this example are proper data anonymization to respect data privacy and centralization to handle data federalism, but if those tasks could be solved this could vastly improve and speed up medical diagnostics.
If you want to contribute you are very welcome! Feel free to send me a [mail](https://www.mathi.uni-heidelberg.de/~pmichl/). If you want to report bugs or request features please feel free to use the [issue tracker](https://github.com/fishroot/nemoa/issues) provided by GitHub. You can also follow the progress of the project by joining the nemoa [google group](http://groups.google.com/group/nemoa) (which is currently not very weel-frequented).
##### Authors ##### Nemoa is maintained by [Patrick Michl](https://www.mathi.uni-heidelberg.de/~pmichl/) with scientific advisory by Prof. Dr. [Rainer König](http://ibios.dkfz.de/tbi/index.php/network-modeling/people/34-koenig)