Woonings, Daniël (2022) On conceptor control in continuous time. Bachelor's Thesis, Artificial Intelligence.
|
Text
Final_draft_DW.pdf Download (1MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (120kB) |
Abstract
Conceptors have been a successful mechanism used in a variety of tasks ranging from machine learning to neuroscience. As of now, they have always been used in discrete-time contexts. This report extends the theory of conceptors to continuous time objects such as the leaky-integrator echo state network and liquid state machine. In particular, the pattern regeneration task is solved for the continuous-time leaky integrator echo state networks using two different methods. The first method is derived from the assumption of a hard conceptor and is based on orthogonal vector projection. The second method is derived for arbitrary conceptors and the negation of a conceptor is used to construct a vector projection. As a consequence of the first method, several properties of hard conceptors are derived and proved. The derived theory is illustrated by two experiments, the first studies pattern regeneration in continuous time. The second experiment uses the conceptor classifier mechanism in combination with a spiking reservoir. Finally, pattern regeneration for spiking reservoirs is discussed
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Abreu, S. and Pourcel, G. A. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 22 Sep 2022 12:40 |
| Last Modified: | 22 Sep 2022 12:41 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/28752 |
Actions (login required)
 |
View Item |