• August 2023: Paper on progressive neural compression accepted for presentation at RTSS 2023

The paper titled “Progressive Neural Compression for Adaptive Image Offloading under Timing Constraints” co-authored with Ruiqi Wang, Hanyang Liu, Jiaming Qiu, Moran Xu, and Chenyang Lu has just been accepted for presentation at RTSS 2023.

The paper considers an image classification system where captured images are uploaded over a variable bandwidth channel and classification decisions need to be made by a certain deadline. The paper proposes a multi-objective rateless autoencoder that optimizes for multiple compression rates via stochastic taildrop to create a compression solution that produces features ordered according to their importance for inference performance, with feature transmissions then proceeding in the order of their importance. The benefits of the approach over state-of-the-art solutions are demonstrated experimentally in an IoT testbed.