| Finding Original Regulatory Networks with Weight Matrices
|
| Nobuhisa UEDA1) and Taisuke SATO1) |
| 1) Dept. of Computer Science, Tokyo Institute of Technology |
| We propose a new method to find regulatory networks with weight matrices from expression patterns. It estimates parameters in the network with a real-coded genetic algorithm called UNDX, finds structures with the random-restart hill-climbing search, and evaluates their fitness with an MDL-based fitness function. We also show experimental results using this method. In experiments, we succeeded in identifying a structure which generated the expression patterns used as data sets, while existing methods failed to discover the original one. | | | |