Advances in Efficient Data Representation and Processing

The field of data representation and processing is moving towards more efficient and scalable solutions. Researchers are exploring novel approaches to compress and store large datasets, such as hypergraphs and sparse matrices, to reduce memory consumption and improve computational efficiency. New algorithms and data structures are being developed to support fast queries and navigation over compressed data. Additionally, there is a growing interest in leveraging advanced networking resources and distributed computing architectures to accelerate data-intensive applications. Notable papers include HybHuff, which presents a hybrid compression framework for hypergraph adjacency formats, and ViFusion, which introduces a communication-aware tensor fusion framework for scalable video feature indexing. Other noteworthy papers are Binsparse, which proposes a cross-platform binary sparse matrix storage format, and LZSE, which develops an LZ-style compressor supporting O(log n)-time random access. PAT and SuperSONIC also present innovative solutions for parallel aggregated trees and cloud-native infrastructure for ML inferencing, respectively.

Sources

HybHuff: Lossless Compression for Hypergraphs via Entropy-Guided Huffman-Bitwise Coordination

ViFusion: In-Network Tensor Fusion for Scalable Video Feature Indexing

Binsparse: A Specification for Cross-Platform Storage of Sparse Matrices and Tensors

LZSE: an LZ-style compressor supporting $O(\log n)$-time random access

PAT: a new algorithm for all-gather and reduce-scatter operations at scale

SuperSONIC: Cloud-Native Infrastructure for ML Inferencing

Succinct Preferential Attachment Graphs

Built with on top of