🕐 2026-04-04 12:04 UTC
📰 パラメータ4個で710M超えのFoundation Modelに勝った時系列予測手法FLAIRの全貌 (👍 35)
🇬🇧 FLAIR beats Amazon's 710M-parameter Chronos model with just 4 parameters. No GPU needed, pure numpy/scipy. Simple yet powerful time series forecast...
🇰🇷 FLAIR가 단 4개 파라미터로 Amazon의 7억1천만 파라미터 Chronos 모델을 능가. GPU 불필요, numpy/scipy만으로 구현된 시계열 예측 기법.
🔗 https://zenn.dev/t_honda/articles/flair-time-series-forecasting
📰 パラメータ4個で710M超えのFoundation Modelに勝った時系列予測手法FLAIRの全貌 (👍 35)
🇬🇧 FLAIR: 4-parameter time series method beats 710M foundation models. Only needs numpy/scipy, no GPU required
🇰🇷 FLAIR: 파라미터 4개로 710M 모델 능가한 시계열 예측 - numpy/scipy만 필요, GPU 불필요
🔗 https://zenn.dev/t_honda/articles/flair-time-series-forecasting
📰 パラメータ4個で710M超えのFoundation Modelに勝った時系列予測手法FLAIRの全貌 (👍 31)
🇬🇧 FLAIR: Just 4 parameters & numpy/scipy outperform Amazon's 710M-param Chronos on 25 time-series benchmarks. No GPU needed, 500-line Python file.
🇰🇷 FLAIR: 단 4개 파라미터와 numpy/scipy로 Amazon의 710M 파라미터 Chronos 능가. GPU 불필요, 500줄 파이썬 파일.
🔗 https://zenn.dev/t_honda/articles/flair-time-series-forecasting
📰 パラメータ4個で710M超えのFoundation Modelに勝った時系列予測手法FLAIRの全貌 (👍 30)
🇬🇧 FLAIR: A 4-parameter time series forecasting method that outperforms Amazon's 710M-parameter Chronos model with just numpy.
🇰🇷 FLAIR: numpy만으로 아마존의 7억 파라미터 모델을 이긴 단 4개 파라미터 시계열 예측 기법.
🔗 https://zenn.dev/t_honda/articles/flair-time-series-forecasting
📰 パラメータ4個で710M超えのFoundation Modelに勝った時系列予測手法FLAIRの全貌 (👍 28)
🇬🇧 FLAIR beats 710M-parameter Chronos with just 4 parameters—pure numpy/scipy, no GPU, outperforming Foundation Models in time series
🇰🇷 FLAIR가 단 4개 파라미터로 710M짜리 Chronos 능가—GPU 없이 numpy/scipy만으로 시계열 예측 최강
🔗 https://zenn.dev/t_honda/articles/flair-time-series-forecasting
🚀 New #rstats 📦: ecoXCorr now available on CRAN!
https://cran.r-project.org/package=ecoXCorr
It provides a simple workflow to explore lagged associations between environmental #timeseries and eco / epidemio responses:
➡️ flexible lag intervals
➡️ GLMM via glmmTMB
➡️ plot cross-correlation maps
see REDME: https://github.com/Nmoiroux/ecoXCorr
🌐🧪🌍
Tools for analysing lagged relationships between environmental variables and ecological or epidemiological time series. The package implements a workflow to aggregate meteorological data over multiple lagged intervals, fit regression models, including mixed-effect models using 'glmmTMB', for each lag window, and visualise varied models outcomes (effect strength and direction, model prediction error...) using cross-correlation maps ('CCM').
📰 パラメータ4個で710M超えのFoundation Modelに勝った時系列予測手法FLAIRの全貌 (👍 22)
🇬🇧 FLAIR time series forecasting: just 4 parameters outperforming Amazon's 710M-parameter Chronos model. No GPU needed, numpy/scipy only.
🇰🇷 FLAIR 시계열 예측: 파라미터 4개로 710M 파라미터 모델 능가. GPU 불필요, numpy/scipy만으로 구현.
🔗 https://zenn.dev/t_honda/articles/flair-time-series-forecasting
Working with time-series data at scale? “How Prometheus Keeps Its TSDB Sane” breaks down how Prometheus keeps its own storage manageable and safe.
Read More: https://zalt.me/blog/2026/04/prometheus-tsdb-sanity
Google's 200M-parameter time-series foundation model with 16k context
https://github.com/google-research/timesfm
#HackerNews #GoogleResearch #TimeSeries #Model #AIInnovation #MachineLearning #ContextualAI
TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting. - google-research/timesfm
1936 #Poland
The speaking clock at the Main Post Office in Kraków, on the corner of Potocki and Wielopole Streets.
A speaking clock is a telephone service that provides the exact current time when you call a designated number.
It was once an essential tool before digital clocks and smartphones became common.
The recorded voice—often of a radio announcer—would announce the time precisely, synchronized with official standards.
Today, most speaking clocks are historical curiosities, but some still operate as reminders of early telecommunication technology.
#telephony #history #ntp #ntps #time #timeseries #phones #telecomunication #krakow #europe
Licence: Public Domain
Sygnatura: 3/1/0/8/5329
Data/Okres powstania: 1936/07
Zenn Trends
Nicolas Moiroux
Mahmoud Zalt
Hacker News
Arkadiusz Sikorski