24 August 2021 in IEEE Transactions on Circuits and Systems for Video Technology
Collaborative Intelligence is a new paradigm that splits a deep neural network (DNN) into an edge and cloud for deploying a DNN-based image recognition application. In this paradigm, deep features, which are the outputs of the edge DNN, are compressed and transmitted to the cloud DNN. Because the deep features have a number of responses that are similar to each other, for efficient compression, previous methods spatially arrange and compress the deep features as an image to utilize the similarity as a spatial correlation. However, if the deep features are arranged in not only spatial but also temporal directions like those in a video, it may be possible to compress them more efficiently by increasing a temporal correlation. To explore this possibility, we propose a “spatio-temporal arrangement”. This method spatially arranges the deep features as images and temporally arranges them as a video with a novel ordering search algorithm. Our method effectively increases the spatial and temporal correlations hidden in the deep features and achieves high compression efficiency compared with the previous methods. Experimental results demonstrate the compression efficiency of our method is better than that of the previous methods (1.50% to 4.98% on BD-Rate evaluation in a lossy setting). Our analysis shows that our method effectively increases the correlation when the input is an image with rich edges and textures.
19 September 2025 in IEEE Access
This paper proposes a fast-parallel method for singular value thresholding, aimed at low-rank analysis of many small matrices. In low-rank analysis, the problem of regularizing the nuclear norm, defined as the sum of singular values, is typically solved by iteratively applying singular value thresholding. However, this process often requires a long run-time due to the computational cost of singular value decomposition. To address this issue, we provide a method that is computationally inexpensive and can be executed in parallel. The core of our approach lies in the discovery of geometric properties of the nuclear norm, leading to a computational process that is fundamentally different from conventional methods. Our method is fast because it does not require singular value decomposition, which has been a bottleneck in terms of computational complexity and parallelism. We design the computation method for two-column matrices as the main proposal, and also demonstrate an optional method for approximately solving the three-column matrices problem. Through experiments, the proposed method calculated singular value thresholding up to 65.81 times faster with higher calculation accuracy than conventional methods.We also applied our method to real data for noise removal in 3D models and confirmed the improvement in computational speed.