<div class="csl-bib-body">
<div class="csl-entry">Furutanpey, A. (2024, June 26). <i>Efficient Generative Modelling for Transmitting Salient Information</i> [Poster Presentation]. Generative Modeling Summer School (GeMSS 2024), Eindhoven, Netherlands (the). http://hdl.handle.net/20.500.12708/205192</div>
</div>
-
dc.identifier.uri
http://hdl.handle.net/20.500.12708/205192
-
dc.description.abstract
We discuss the challenges of efficient data transmission and recovery in modern communications systems, focusing on edge-cloud computing environments with Deep Neural Networks (DNNs) and the role of generative models. We present our previous work on Shallow Variational Bottleneck Injection (SVBI) as a novel approach to optimize bandwidth usage while maintaining prediction accuracy. The study examines three fundamental challenges: the efficient utilization of mobile-grade resources for data compression versus DNN computation, the reliable recovery of human-interpretable representations from compressed signals, and improved ultra-low bitrate perceptual codecs' reliability. The research is particularly relevant in two contrasting scenarios: urban environments with sophisticated 5G networks and remote environments like Low Earth Orbit satellites where bandwidth and computational resources are severely constrained. The work demonstrates that our methods can effectively preserve prediction integrity while minimizing bandwidth usage. Moreover, we demonstrate that we can reliably recover the original input signal with high perceptual fidelity, which is especially crucial for applications requiring expert review or intervention. The proposed solutions address the increasing tension between the rise of foundational AI models and the limitations of mobile AI accelerators with real-world applicability. Our results demonstrate that the proposed methods are significant advancements for efficient data transmission in resource-constrained environments.
en
dc.language.iso
en
-
dc.subject
Generative Modeling
en
dc.subject
Neural Compression
en
dc.subject
Edge Computing
en
dc.title
Efficient Generative Modelling for Transmitting Salient Information
en
dc.type
Presentation
en
dc.type
Vortrag
de
dc.type.category
Poster Presentation
-
tuw.researchTopic.id
I4
-
tuw.researchTopic.name
Information Systems Engineering
-
tuw.researchTopic.value
100
-
tuw.publication.orgunit
E194-02 - Forschungsbereich Distributed Systems
-
tuw.author.orcid
0000-0001-5621-7899
-
tuw.event.name
Generative Modeling Summer School (GeMSS 2024)
en
tuw.event.startdate
24-06-2024
-
tuw.event.enddate
28-06-2024
-
tuw.event.online
On Site
-
tuw.event.type
Event for scientific audience
-
tuw.event.place
Eindhoven
-
tuw.event.country
NL
-
tuw.event.presenter
Furutanpey, Alireza
-
wb.sciencebranch
Informatik
-
wb.sciencebranch.oefos
1020
-
wb.sciencebranch.value
100
-
item.languageiso639-1
en
-
item.openairetype
conference poster not in proceedings
-
item.grantfulltext
none
-
item.fulltext
no Fulltext
-
item.cerifentitytype
Publications
-
item.openairecristype
http://purl.org/coar/resource_type/c_18co
-
crisitem.author.dept
E194-02 - Forschungsbereich Distributed Systems
-
crisitem.author.orcid
0000-0001-5621-7899
-
crisitem.author.parentorg
E194 - Institut für Information Systems Engineering
