Deep Learning as a Service (DLaaS)
Tooling to improve the usability and efficiency of deep learning for third parties.
Several demonstrators will show how to exploit DL as a Service.
DurationJanuary 2019 - December 2023
We live in an era where the amount of data being collected is truly overwhelming. Deep learning (DL) techniques can help greatly to exploit this wealth of data. A variety of packages and services are available that make DL readily available to non-experts. Examples of software packages are TensorFlow, MXNet, and Caffe. Services that on top provide server capacity to run the DL algorithms are for instance Amazon Web Services, Microsoft Azure, and Google Cloud.
The project addresses four main data efficient DL approaches:
• It develops a more easily usable service for applying deep learning as an extension to existing libraries that is able to run DL approaches more efficiently.
• The service in addition offers efficient DL approaches to cope with a lack of labels.
• The approaches allow users to gain insights into the DL networks and configurations.
• The service embeds DL approaches that can cope better with temporal high frequency data.
The related award: TTW P16-25 - Project 1.
Sander Stuijk is project leader for the TUe part of this project.
Researchers involved in this project
This project is part of the EDL program and has received funding from NWO-TTW with the grant number P16-25, Project 1.