|
Transparent modelling of uncertainty - a role for fuzzy sets and systems in XAI?| old_uid | 17735 |
|---|
| title | Transparent modelling of uncertainty - a role for fuzzy sets and systems in XAI? |
|---|
| start_date | 2019/06/06 |
|---|
| schedule | 14h |
|---|
| online | no |
|---|
| location_info | 24-25, 4e étage, salle 405 |
|---|
| summary | While fuzzy sets and systems are often credited with being useful for modelling uncertainty and for being interpretable, it is straightforward to argue that they are neither. With both uncertainty handling and interpretability being 'hot topics' in AI, in this talk I reflect on the capacity of fuzzy sets and systems to transparently model and handle uncertainty, focussing in particular on data-driven fuzzy set design and novel approaches to non-singleton fuzzy logic systems.
In respect to the latter, I will discuss their utility in respect to 'performance' but also more generally in respect to the important reason they play in the systematic and transparent modelling of uncertainty affecting system inputs. Finally, time permitting, I will focus on how to capture uncertainty in the real world, presenting recent interdisciplinary work at the interface of computer and the social sciences. |
|---|
| responsibles | Piwowarski |
|---|
| |
|