Universal Representations for Classification-enhanced Lossy Compression
This work addresses the challenge of efficient encoder design in compression systems for scenarios requiring flexibility in tradeoffs, though it is incremental as it builds on existing rate-distortion-classification frameworks.
The paper tackles the problem of designing a single encoder for lossy compression that works across multiple decoding objectives, such as distortion and classification accuracy, avoiding the need for retraining for each specific tradeoff. Experimental results on MNIST show minimal performance degradation for perceptual tasks but a significant distortion penalty when reusing an encoder optimized for one classification-distortion tradeoff in other settings.
In lossy compression, the classical tradeoff between compression rate and reconstruction distortion has traditionally guided algorithm design. However, Blau and Michaeli [5] introduced a generalized framework, known as the rate-distortion-perception (RDP) function, incorporating perceptual quality as an additional dimension of evaluation. More recently, the rate-distortion-classification (RDC) function was investigated in [19], evaluating compression performance by considering classification accuracy alongside distortion. In this paper, we explore universal representations, where a single encoder is developed to achieve multiple decoding objectives across various distortion and classification (or perception) constraints. This universality avoids retraining encoders for each specific operating point within these tradeoffs. Our experimental validation on the MNIST dataset indicates that a universal encoder incurs only minimal performance degradation compared to individually optimized encoders for perceptual image compression tasks, aligning with prior results from [23]. Nonetheless, we also identify that in the RDC setting, reusing an encoder optimized for one specific classification-distortion tradeoff leads to a significant distortion penalty when applied to alternative points.