I gave a chat within the workshop on how the synthesis of logic and device Mastering, In particular places like statistical relational Discovering, can allow interpretability.
Weighted design counting typically assumes that weights are only specified on literals, typically necessitating the need to introduce auxillary variables. We take into account a whole new technique based upon psuedo-Boolean capabilities, resulting in a more common definition. Empirically, we also get SOTA results.
I gave a talk entitled "Perspectives on Explainable AI," at an interdisciplinary workshop focusing on developing believe in in AI.
I attended the SML workshop during the Black Forest, and discussed the connections between explainable AI and statistical relational Finding out.
Our paper (joint with Amelie Levray) on Understanding credal sum-product or service networks is approved to AKBC. These networks, together with other kinds of probabilistic circuits, are beautiful simply because they assurance that sure kinds of probability estimation queries could be computed in time linear in the dimensions on the network.
The article, to seem inside the Biochemist, surveys a number of the motivations and approaches for building AI interpretable and accountable.
The function is enthusiastic by the need to take a look at and Assess inference algorithms. A combinatorial argument to the correctness of the Tips is usually deemed. Preprint below.
A journal paper has been approved on prior constraints in tractable probabilistic types, readily available to the papers tab. Congratulations Giannis!
A new collaboration Along with the NatWest Group on explainable equipment Understanding is discussed from the Scotsman. Website link to article below. A preprint on the outcomes will be designed accessible Soon.
Along with colleagues from Edinburgh and Herriot Watt, We've place out the call for a different research agenda.
Paulius' work on algorithmic methods for randomly producing logic plans and probabilistic logic plans has become acknowledged for the principles and practise of constraint programming https://vaishakbelle.com/ (CP2020).
Our MLJ (2017) write-up on scheduling with hybrid MDPs was acknowledged for presentation in the journal observe.
Our work on synthesizing options with loops during the presence of noise will seem inside the Global journal of approximate reasoning.
Our get the job done (with Giannis) surveying and distilling approaches to explainability in machine Understanding is acknowledged. Preprint right here, but the final Variation are going to be on the web and open up entry before long.