Atsuki Sato

I am a first-year Ph.D. student at the Matsui Lab in the Graduate School of Information Science and Technology, The University of Tokyo.
My research focuses on developing learning-augmented data structures and algorithms, aiming to enhance computational efficiency and provide robust theoretical guarantees.

Email / Google Scholar / Github

Publications

	Fast Partitioned Learned Bloom Filter Atsuki Sato, Yusuke Matsui NeurIPS 2023 (Poster), Code / arXiv We propose Fast PLBF and Fast PLBF++, two methods that significantly reduce the construction time of Partitioned Learned Bloom Filter while maintaining the excellent memory efficiency.
	PCF Learned Sort: a Learning Augmented Sort Algorithm with O(nloglogn) Expected Complexity Atsuki Sato, Yusuke Matsui arXiv 2024, arXiv We propose PCF Learned Sort, the first learning-augmented sort algorithm with provable O(nloglogn) expected complexity under mild distributional assumptions.
	Fast Construction of Partitioned Learned Bloom Filter with Theoretical Guarantees Atsuki Sato, Yusuke Matsui arXiv 2024, arXiv We propose fast PLBF, fast PLBF++, and fast PLBF#, which significantly reduce the original PLBF's O(N³k) construction time to O(N²k), O(NklogN), and O(Nklogk), respectively.
	Cascaded Learned Bloom Filter for Optimal Model-Filter Size Balance and Fast Rejection Atsuki Sato, Yusuke Matsui arXiv 2025, arXiv We propose CLBF, a cascaded learned Bloom filter that optimally balances model and filter sizes while minimizing reject time, achieving up to 24% memory savings and 14x faster rejection.
	Media of Langue: Exploring Word Translation Network Goki Muramoto, Atsuki Sato, Takayoshi Koyama NAACL 2025 (Findings), Website / arXiv We discover the massive network formed by chains of translations as Word Translation Network, and propose Word Translation Map as a novel interface for exploring this network.