โจ We kicked off a new season of talks, the AI FUN with ELLIS Invited Speaker Lecture series, brought to you by the Centre for AI FUNdamentals and the #ELLIS Unit at Manchester ๐
The 1๏ธโฃst talk was given by our guest speaker Rafael Izbicki (@rafael-izbicki.bsky.social) from UFSCar in Brazil ๐ง๐ท
๐ฆ Our new paper is out in PLOS Neglected Tropical Diseases! We use Google Trends + surveillance data to improve nowcasting of dengue in Brazil.
W/ the great Y. Xiao, G. Soares, L. Bastos & P. Moraga.
๐ journals.plos.org/plosntds/art...
Online Tracker: diseasesurveillance.github.io/dengue-track...
Interested in uncertainty quantification methods such as conformal prediction, Gaussian processes and conditional density estimators?
You can now get a physical copy of my book on Amazon!
amazon.com/Machine-Lear...
#MachineLearning #UncertaintyQuantification #AI #DataScience #ConformalPrediction
๐ข New on Medium: My Paper Was Outdated Before It Was Published
Submitted. Waited. Rejected. Revised. Accepted.
By then, 50+ citations and 1k GitHub stars. The field had moved on.
Thoughts on fixing peer review in fast-moving fields:
๐ rizbicki.medium.com/my-paper-was...
#PeerReview
What if you could beat the Likelihood Ratio Test?
A fun and potentially transformational phy-statistics paper with Ann Lee, James Carzon, @rafael-izbicki.bsky.social, luca Masserano and @danielwhiteson.bsky.social
arxiv.org/abs/2507.17831
Big newsโour paper was accepted for oral presentation at #UAI2025! ๐ฅณ
We (me + Vagner S. Santos, Thiago R Ramos, @rafael-izbicki.bsky.social ) built EPICSCORE, a method to improve conformal prediction intervals by adding epistemic uncertainty awareness.
Hereโs why it matters ๐งต
Interesed in uncertainty quantification methods such as conformal prediction, Gaussian processes, and conditional density estimators?
You can now get a physical copy of my book on Amazon!
๐ amazon.com/Machine-Lear...
#MachineLearning #UncertaintyQuantification #AI
Interesed in uncertainty quantification methods such as conformal prediction, Gaussian processes, and conditional density estimators?
You can now get a physical copy of my book on Amazon!
๐ amazon.com/Machine-Lear...
#MachineLearning #UncertaintyQuantification #AI
Aparentemente esse ano serรก um caos em Sรฃo Paulo em relaรงรฃo ร dengue novamente.
Mais estimativas em diseasesurveillance.github.io/dengue-track...
#dengue #epi #epidemiology
No, it doesn't work. For multiclass you need to direcly change the loss function (see e.g. Eq 2.3 of rafaelizbicki.com/UQ4ML.pdf)
[Se vocรช estรก no Brasil, acesse rafaelizbicki.com/UQ4MLpt ]
Topics include conditional density estimation, Gaussian processes, conformal prediction, and calibration techniques.
Download for free or buy printed version
๐http://rafaelizbicki.com/UQ4ML
#MachineLearning #UncertaintyQuantification #DataScience #AI #PredictiveAnalytics
๐ฎ New Book Release! ๐
Happy to share my new book, Machine Learning Beyond Point Predictions: Uncertainty Quantification.
Tools to tackle aleatoric & epistemic uncertainties in AI. +
ร absurdo como muitos dos divulgadores cientรญficos simplesmente pararam de cobrar vacinas quando o novo governo assumiu.
Estamos faz meses tentando dar uma dose de vacina de covid para minha filha, mas estรก sempre em falta. Idem para idosos.
๐ Check it out: sciencedirect.com/science/arti...
#ClimateScience #UncertaintyQuantification #DataScience
๐ณ New paper! Led by Denis Valle, we present a nonparametric approach to quantify uncertainty in multistep upscaling models. We illustrate our methods for forest biomass predictions in the Brazilian Amazon. +
#RemoteSensing #MachineLearning #ForestBiomass
Decision regions based on a significance level alpha and a threshold epsilon for the gender wage gap data presented in the paper. For higher values of alpha and lower values of epsilon, the hypothesis is rejected. For lower values of alpha and higher values of epsilon, the hypothesis is not rejected.
I am happy to announce that a paper of mine, Rafael B. Stern and @rafael-izbicki.bsky.social has been accepted for publication at the International Journal of Approximate Reasoning ๐
In our paper, we show how to build and test hypotheses of practical equivalence in nonparametric settings. ๐งต
We demonstrate the capabilities of LF2I in a ๐ก๐ข๐ ๐ก-๐๐ง๐๐ซ๐ ๐ฒ ๐ฉ๐ก๐ฒ๐ฌ๐ข๐๐ฌ ๐๐ฉ๐ฉ๐ฅ๐ข๐๐๐ญ๐ข๐จ๐ง, addressing challenges involving intractable likelihoods and high-dimensional datasets.โฃ
โฃ
๐ Read the full article here: lnkd.in/dxRgHZep
โฃ
#Statistics #MachineLearning #DataScience
In this work, we introduce ๐๐ ๐๐, a modular framework that combines tools from classical statistics and machine learning to construct confidence sets for problems where statistical models are given by ๐๐จ๐ฆ๐ฉ๐ฅ๐๐ฑ ๐๐จ๐ซ๐ฐ๐๐ซ๐ ๐ฌ๐ข๐ฆ๐ฎ๐ฅ๐๐ญ๐จ๐ซ๐ฌ (LFI) +โฃ
Excited to announce our new publication!
Our paper, "Likelihood-free frequentist inference: bridging classical statistics and machine learning for reliable simulator-based inference," is now in EJS!
Co-authored with the great N Dalmasso, @lucamasse95
, D Zhao, and A B Lee. +
No dia 25/11, eu, Thiago Rodrigo Ramos e Luben Miguel Cruz Cabezas daremos uma sรฉrie de seminรกrios no Programa de Pรณs-Graduaรงรฃo do Dept. de Ginecologia e Obstetrรญcia da FMRP USP. Mais informaรงรตes: lnkd.in/dps4mHDU
Stop oversampling! Changing the cutoff in probabilistic classifiers is enough for imbalanced data.
In our new paper, Gabriel O. Assunรงรฃo, Marcos O. Prates, and I explore this in depth. jds-online.org/journal/JDS/...
#DataScience #MachineLearning #ImbalancedData #AI #Oversampling
๐ข Tarde Temรกtica: Inferรชncia, Redes e IA no dia 25/10 no CINA-UFSCar!
Palestras com @franciscorodrigues.bsky.social (ICMC-USP), Pedro L. C. Rodrigues (Inria) e Heloรญsa Camargo (DC-UFSCar).
Saiba mais: rafaelizbicki.com/tematico_ia/
#InteligenciaArtificial #Redes #Simulaรงรฃo #DataScience
Tambรฉm estou com a sensaรงรฃo de fim do mundo
Tรก insuportรกvel isso aqui. O mais triste รฉ que passaram a semana dizendo que hoje ia esfriar um monte.
Ran my text through ChatGPT arguing why oversampling usually doesnโt make sense, and it flipped my argument to support oversampling instead. A reminder that AI often defaults to consensusโeven when it's wrong. Always double-check!
#AI #MachineLearning #Statistics
Does not exist
Estรฃo abertas as inscriรงรตes para o Concurso Pรบblico para Professor Efetivo no Depto de Estatรญstica da UFSCar! Inscriรงรตes atรฉ 11/10/2024
Sรฃo Carlos รฉ um polo de inovaรงรฃo e pesquisa, com excelente qualidade de vida. + info: concursos.ufscar.br
#concurso #estatistica #datascience #machinelearning
Adoro esse livro!
O PIPGEs (UFSCar/USP) estรก com inscriรงรตes abertas para mestrado (atรฉ 21/10) e doutorado (atรฉ 14/10). Bolsas disponรญveis e oportunidades de tutoria e monitoria. Linhas de pesquisa: Machine Learning, Inferรชncia, Probabilidade e Modelos de Regressรฃo. + info: tinyurl.com/pipges2024 #posgrad
