This document discusses optimizations for deep learning frameworks on Intel CPUs and Fugaku processors. It introduces oneDNN, an Intel performance library for deep neural networks. JIT assembly using Xbyak is proposed to generate optimized code depending on parameters at runtime. Xbyak has been extended to AArch64 as Xbyak_aarch64 to support Fugaku. AVX-512 SIMD instructions are briefly explained.
技術動向の調査として、ICML Workshop Uncertainty & Robustness in Deep Learningの中で、面白そうなタイトルを中心に読んで各論文を4スライドでまとめました。
最新版:https://speakerdeck.com/masatoto/icml-2021-workshop-shen-ceng-xue-xi-falsebu-que-shi-xing-nituite-e0debbd2-62a7-4922-a809-cb07c5da2d08(文章を修正しました。)
Topological Data Analysis of Complex Spatial SystemsMason Porter
?
Topological data analysis is a method for studying complex systems and high-dimensional data by examining the "shape" of data using techniques from computational topology like persistent homology. The document discusses applications of topological data analysis to spatial networks, spider webs, voting data, and COVID-19 case data. It also compares different methods for constructing simplicial complexes from data for use in persistent homology calculations.
The document contains contact information for Ichigaku Takigawa including their email address ichigaku.takigawa@riken.jp, personal website URL https://itakigawa.github.io/, and mentions they are working with IBISML and ATR on materials informatics and bioinformatics. It also includes a link to their page https://itakigawa.page.link/IBISML for a PDF document.
This document discusses optimizations for deep learning frameworks on Intel CPUs and Fugaku processors. It introduces oneDNN, an Intel performance library for deep neural networks. JIT assembly using Xbyak is proposed to generate optimized code depending on parameters at runtime. Xbyak has been extended to AArch64 as Xbyak_aarch64 to support Fugaku. AVX-512 SIMD instructions are briefly explained.
技術動向の調査として、ICML Workshop Uncertainty & Robustness in Deep Learningの中で、面白そうなタイトルを中心に読んで各論文を4スライドでまとめました。
最新版:https://speakerdeck.com/masatoto/icml-2021-workshop-shen-ceng-xue-xi-falsebu-que-shi-xing-nituite-e0debbd2-62a7-4922-a809-cb07c5da2d08(文章を修正しました。)
Topological Data Analysis of Complex Spatial SystemsMason Porter
?
Topological data analysis is a method for studying complex systems and high-dimensional data by examining the "shape" of data using techniques from computational topology like persistent homology. The document discusses applications of topological data analysis to spatial networks, spider webs, voting data, and COVID-19 case data. It also compares different methods for constructing simplicial complexes from data for use in persistent homology calculations.
The document contains contact information for Ichigaku Takigawa including their email address ichigaku.takigawa@riken.jp, personal website URL https://itakigawa.github.io/, and mentions they are working with IBISML and ATR on materials informatics and bioinformatics. It also includes a link to their page https://itakigawa.page.link/IBISML for a PDF document.