Kimon Fountoulakis (@kfountou)
GPT-5.2가 COLT 2022의 오픈 문제를 해결했다고 보고하는 논문(버전 2)을 공개했습니다. 하나의 그래디언트만 사용하는 가속화 방법으로 L1-정규화된 PageRank에 대한 전체 작업량 경계를 증명하고, 일부 경우 비가속 방법보다 우수함을 보였다고 설명합니다.
GPT-5.2 solves our COLT 2022 open problem using an accelerated method with one gradient per iteration (Version 2, the paper). Bounding the total work of accelerated methods for L1-regularized PageRank, and showing that it can be better than non-accelerated methods in some
Как я построил Graph RAG систему с точностью 96.7% за 5 дней: от научных статей до production-ready пайплайна
Я реализовал Graph RAG систему, которая комбинирует 5 техник из свежих научных статей (KET-RAG, HippoRAG 2, VectorCypher) в единый пайплайн с декларативным Datalog reasoning-движком, полной провенансной трассировкой и типизированным API. Результат: 174/180 (96.7%) на билингвальном бенчмарке из 30 вопросов, оценённых в 6 режимах retrieval. Три режима достигли 100%. В статье — архитектура, 10 уроков оптимизации и эволюция от 38% до 96.7% за 10 итераций.
https://habr.com/ru/articles/1003064/
#GraphRAG #RAG #Neo4j #NLP #LLM #Python #Datalog #Knowledge_Graph #embeddings #PageRank
Skeleton Indexing (KDD 2025) + HippoRAG 2 (ICML 2025) + VectorCypher + Datalog Reasoning + 10 итераций оптимизации TL;DR Я реализовал Graph RAG систему, которая комбинирует 5 техник из свежих научных...
In his 1948 paper "A Mathematical Theory of Communication", C. E. Shannon put aside the issue of the meaning of what is communicated: "These semantic aspects of communication are irrelevant to the engineering problem." That is, Shannon's "information" is separated from content and hence also from truth. The irrelevance of
My favourite things to teach in #Geography and #GIS is how timely and relevant the field is.
Case in point:
On today's lecture about network analysis, we discussed how travel time analyses can be fudged to make one mode of transit appear better/worse than it is, and how this could also be unintentional. The vitality and attractiveness of the #Helsinki urban centre is under constant public debate in the media from this perspective.
Yeah, and there is the #PageRank thing as well.
#Google's #PageRank #Algorithm is a core component of its search engine, designed to measure the importance and authority of web pages based on the quantity and quality of links pointing to them.
Developed by Google co-founders Larry Page and Sergey Brin at Stanford University, the algorithm treats links as "votes" of confidence, where votes from more important pages weigh more heavily.
It is interesting to ponder the similarities and differences of the classes of #blockchain and #llm algorithms and how they have been embraced by different actors.
There is an obvious manic search for the next overwhelming digital monopoly that will be built on exploiting some algorithmic advance.
This goes back to the success of Google's #pagerank algorithm but there are also earlier echoes in algorithms used in quantitative finance (and a previous bubble, now almost entirely forgotten)
1/2
sayzard
Habr
Andrew Shields
Tuomas Väisänen 📼🧟♂️
Knowledge Zone
PPC Land
Open Risk
Information
gary