Goal-oriented communication achieves high accuracy with minimal data
Goal-Oriented Ordered Autoencoding for Successive Refinement in Progressive Communication
Information Theory
Summary
Communication systems often struggle to keep tasks accurate when the amount of data sent changes unpredictably. The authors developed a method called GO-OAE that sends information in an order where the receiver can get better results as more data arrives. Instead of focusing on sending perfect copies, it focuses on sending what is important for a task, like recognizing images. Their tests show this method can compress images significantly while still keeping high accuracy in recognition.
What this means in practice
- •For mobile app developers: Create apps that send progressively improving compressed image features for recognition under changing network bandwidth.
- •For internet of things engineers: Enable IoT devices to send minimal coded information for task-specific inference that improves as more data arrives.
Authors
Qi Liao, Dario Bega
Abstract
Goal-oriented communication often struggles to sustain downstream task performance under variable and unpredictable bandwidth. We address this problem by proposing GO-OAE (Goal-Oriented Ordered AutoEncoding), which learns a progressively refinable quantized latent sequence, optimized directly for task utility rather than source reconstruction. Each component of the sequence is represented for transmission by its code index, and the learned ordering enables inference from any received prefix of the index sequence, with performance improving as more indices arrive. We prove that optimizing the goal-oriented distortion objective is equivalent to solving a weighted conditional information bottleneck formulation, and introduce an unbiased multi-prefix importance sampling (MPIS) estimator, to reduce GO-OAE training computational cost from O(BK) to O(Bm) with m << K, and reduces the estimator variance by a factor of 1/m. Experiments show that, on MiniImageNet, GO-OAE achieves over 96% of the lossless upper-bound classification accuracy using 324 bits per image (683X compression vs. raw RGB), and already attains over 94% of the upper bound with only 70% of the transmitted indices. It consistently outperforms reconstruction-optimized codecs and ordered autoencoders trained with uniform prefix sampling across all transmission budgets, demonstrating effective communication at extreme compression rates.