Network quantization has proven to be a powerful approach to reduce the memory and computational demands of deep learning models for deployment on resource-constrained devices. However, traditional quantization methods often rely on access to training data, which is impractical in many real-world scenarios due to privacy, security, or regulatory constraints. Zero-shot Quantization (ZSQ) emerges as a promising solution, achieving quantization without requiring any real data. In this paper, we provide a comprehensive overview of ZSQ methods and their recent advancements. First, we provide a formal definition of the ZSQ problem and highlight the key challenges. Then, we categorize the existing ZSQ methods into classes based on data generation strategies, and analyze their motivations, core ideas, and key takeaways. Lastly, we suggest future research directions to address the remaining limitations and advance the field of ZSQ. To the best of our knowledge, this paper is the first in-depth survey on ZSQ.

BACKGROUND: As next-generation sequencing technology made rapid and cost-effective sequencing available, the importance of computational approaches in finding and analyzing copy number variations (CNVs) has been amplified. Furthermore, most genome projects need to accurately analyze sequences with fairly low-coverage read data. It is urgently needed to develop a method to detect the exact types and locations of CNVs from low coverage read data. RESULTS: Here, we propose a new CNV detection method, CNV_SS, which uses scale-space filtering. The scale-space filtering is evaluated by applying to the read coverage data the Gaussian convolution for various scales according to a given scaling parameter. Next, by differentiating twice and finding zero-crossing points, inflection points of scale-space filtered read coverage data are calculated per scale. Then, the types and the exact locations of CNVs are obtained by analyzing the finger print map, the contours of zero-crossing points for various scales. CONCLUSIONS: The performance of CNV_SS showed that FNR and FPR stay in the range of 1.27% to 2.43% and 1.14% to 2.44%, respectively, even at a relatively low coverage (0.5x ≤C ≤2x). CNV_SS gave also much more effective results than the conventional methods in the evaluation of FNR, at 3.82% at least and 76.97% at most even when the coverage level of read data is low. CNV_SS source code is freely available from http://dblab.hallym.ac.kr/CNV SS/.

UNLABELLED: FX is an RNA-Seq analysis tool, which runs in parallel on cloud computing infrastructure, for the estimation of gene expression levels and genomic variant calling. In the mapping of short RNA-Seq reads, FX uses a transcriptome-based reference primarily, generated from ~160 000 mRNA sequences from RefSeq, UCSC and Ensembl databases. This approach reduces the misalignment of reads originating from splicing junctions. Unmapped reads not aligned on known transcripts are then mapped on the human genome reference. FX allows analysis of RNA-Seq data on cloud computing infrastructures, supporting access through a user-friendly web interface. AVAILABILITY: FX is freely available on the web at (http://fx.gmi.ac.kr), and can be installed on local Hadoop clusters. Guidance for the installation and operation of FX can be found under the 'Documentation' menu on the website. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.