STRF estimates prediction models for spatio-temporal data. The parameter storage is compressed by removing uninformative parameters in a systematic way. For finding the best parameters via maximum likelihood estimation, a separable optimization algorithm that can be performed independently in parallel in each graph node is implemented in STRF.

The software is released under Apache License, Version 2.0 and was implemented within the experiments of the article Spatio-temporal random fields: compressible representation and distributed estimation .

**Results that are based on this software will be presented at ECML PKDD 2013 in the talk "Spatio-Temporal Random Fields: Compressible Representation and Distributed Estimation"!**

The following will train a model on a sample of the temperature data with a setup like the one from the paper.

- Download and extract the binary version of STRF.
- Change to the directory and execute
`./strf --train temp1 --test temp2 --time 48 --lambda1 0.0000001 --lambda2 0.0000000001 --ls_max 0.033`. - In STRF, select [File]->[Open] from the menu and open
`TEMPERATURE.DB`from the`graphs`folder (contained in the binary version download). - The empirical temporal dynamics of each vertex may be visualized by right click-ing a vertex and selecting [Plot]->[Empirical] from the pop-up menu and selecting the [Temporal Dynamics] tab. Or select [Plot]->[Current estimate] to plot the corresponding estimate (initially, this is the uniform distribution).
- To start learning: Select [Compute]->[Optimize] from the menu and choose 10 iterations.
- After learning, the new Current estimated vertex marginals may be added to the plot (as descrived above).

A more detailed description will be added soon.

The binary version of STRF that can be downloaded below, requires GTK3 to be installed on your system. Ubuntu Linux 13.04 is known to work with this version.

To build STRF,

- download the compressed source code from the link below,
- decompress the file to a directory of your choice,
- change to this directory and execute the
`./build`command.

**Please notice**: for now, only linux with g++-4.4 is supported. Furthermore, it is required that the libraries

- GTK3
- MySQL Client
- SQLite3
- xerces-c 3.1
- boost program options >=1.42

are installed on your system.

The raw temperature data can be downloaded from Intel Berkeley Research lab. The binary version of STRF includes this dataset already.

Downloads for the traffic data set will be added soon.

**Source code and binary versions for linux and windows, without the need for any additional library, will be available before at the end of august 2013!**

SFB876-C1