|
| Name |
 |
|
|
|
|
Sicco Verwer
|
|
|
| RTI |
|
|
|
|
|
Look at the Projects section for the code of my algorithm for learning real-time automata unsupervised from unlabeled data!
I recently updated it, fixed some minor bugs and added a search routine.
|
|
|
| Interests |
|
|
|
|
|
My interests lie in the fields of learning theory, grammatical inference, complexity theory, and information theory. My main goal is to write efficient algorithms for the identification of (learning) finite state machines. I want to identify the entire structure of such a machine, not just its parameters given the structure.
|
|
|
|
I try to identify machines known as timed automata. These automata include timing relations explicitly, i.e. using numbers. The reasons for identifying these models are twofold: Firstly, they can model systems in an intuitively appealing way. Secondly, we believe that identifyig such models from timed data is easier (i.e. more efficient) then identifying a model that models time implicitly (such as deterministic finite state automata and hidden Markov models).
|
|
|
|
My research focusses on the complexity of identifying and teaching the class of timed automata. Due to negative results regarding these complexities, we have written an algorithm for identifying a simple type of timed automaton, known as a real-time automaton. We have shown experimentally that this algorithm outperforms a similar method that idenitifies a deterministic finite state automaton from the same data when sampling at some fixed frequency.
|
|
|
| Position |
|
|
|
|
|
PhD. student with the Algorithmics Group of
the TU Delft, Faculty of Engineering,
Mathematics and Computer Science (EWI).
|
|
|
| Contact |
|
|
|
|
|
|
|
|
|
|
Home
