BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Women@CL Talklet Event - Lucie Charlotte Magister and Zahra Tarkha
 ni and Tong Xia
DTSTART:20211104T130000Z
DTEND:20211104T140000Z
UID:TALK165256@talks.cam.ac.uk
CONTACT:Lorena Qendro
DESCRIPTION:*Zoom link*: https://cl-cam-ac-uk.zoom.us/j/93420534307?pwd=dT
 g2OXRidXZNdTdhSmhscGpJck40dz09\n\n----------------------------------------
 ----\n\n*Speaker*: Lucie Charlotte Magister\n\n*Title*: GCExplainer: Conce
 pt-based Explanations for Graph Neural Networks\n\n*Abstract*:\nWhile grap
 h neural networks (GNNs) have been shown to perform well on graph-based da
 ta from a variety of fields\, they suffer from a lack of transparency and 
 accountability\, which hinders trust and consequently the deployment of su
 ch models in high-stake and safety-critical scenarios. Even though recent 
 research has investigated methods for explaining GNNs\, these methods are 
 limited to single-instance explanations\, also known as local explanations
 . In this talk\, I will present GCExplainer\, an unsupervised approach for
  post-hoc discovery and extraction of global concept-based explanations fo
 r GNNs. \n\n--------------------------------------------\n\n*Speaker*: Zah
 ra Tarkhani\n\n*Title*: Effective Attack Investigation Across Heterogeneou
 s Isolation Boundaries\n\n*Abstract*: Recent improvements in hardware isol
 ation support have led to numerous new forms of application compartmentali
 sation\, including intra-address space and TEE/enclave-based mechanisms.\n
 These systems aim to reduce the software attack surface by dividing it int
 o isolated components that communicate using well-defined channels. Howeve
 r\, many vulnerabilities inside and across such heterogenous isolation bou
 ndaries cannot easily be detected or debugged. This talk will propose an e
 xtensible framework for investigating attacks in compartmentalised applica
 tions across various isolation boundaries such as processes\, intra-proces
 s sandboxes\, and enclaves.\nHence developers could analyze\, debug\, and 
 audit a wide range of non-trivial threats (e.g.\, memory or synchronizatio
 n vulnerabilities) across compartments to ensure safe interactions\, data/
 resource sharing\, or compartment migrations.\n\n-------------------------
 -------------------\n\n*Speaker*: Tong Xia\n\n*Title*: Uncertain-aware Mac
 hine Learning for Biosignal-based Healthcare Applications\n\n*Abstract*: A
  biosignal is a signal in human beings that can be continually measured li
 ke respiratory sound\, heart activity (ECG)\, brain waves (EEG)\, etc. Bas
 ed on that\, a number of machine learning models have been developed for d
 isease detection and health status monitoring. Although those models yield
  promising accuracy\, because of the lack of interpretation and the usual 
 over-confident predictions\, it is non-trivial to gain people’s trust fo
 r their real deployment in the healthcare field. Predictive uncertainty pr
 ovides the opportunity to inspect a model’s confidence toward its output
 \, thus enabling the less trustworthy predictions to be further checked by
  medical professions. In this talk\, I will reflect on how to incorporate 
 uncertainty estimation to improve the decision-making of ML models for hea
 lthcare.\n\n--------------------------------------------\n\n
LOCATION:Computer Laboratory\, William Gates Building\, FW11 And Remote
END:VEVENT
END:VCALENDAR