Lucas Resende

Lucas Resende

Ph.D. Candidate in Probability at IMPA.
lucas.resende [at]

I'm broadly interested in High Dimensional Probability and its
applications to Mathematical Statistics and Machine Learning.
Currently working on Robust Statistics.

Please feel free to get in touch!


Trimmed sample means for robust uniform mean estimation and regression

An statistical method is robust when it performs well under weak assumptions on the data, such as allowing heavy-tailed distributions or even contaminated data. In this work we use the trimmed mean to obtain robust estimators for uniform mean and regression, obtaining optimal performance in terms of moments.

View in Arxiv Poster

Deep Hashing via Householder Quantization

Most hashing aproaches based on Deep Learning rely on minimizing some similarity loss with a penalty term to force quantization. Observing that typical similarity losses are invariant by rotation we propose to split the learning process in two stages: first learn the similarity without penalization and then learn a rotation that provides good quantization error.

View in Arxiv

Contributions to Dynamic Analysis of Differential Evolution Algorithms

The Differential Evolution is an evolutionary optimization algorithm. In this work new insights are provided on the high-dimensional behavior of the algorithm, including the existence of fixed points and the justification of some well-known rules of thumb for hyper-parameter selection.

View in Evolutionary Computation

Quantifying protocols for safe school activities

This publication is part a multidisciplinary effort to understand the risks associated with reopening schools during the pandemic. Specifically, my work focuses on the reconstruction of unreported and missing data by utilizing ICU and death information, enabling the estimation, at any given time, of the number of individuals in each SEIR state.

View in Plos One

Modeling Geospatial Uncertainty of Geometallurgical Variables with Bayesian Models and Hilbert–Kriging

In mine planning, geospatial estimates of certain variables are crucial for minimizing energy consumption at the plant while maximizing mineral recovery. In this work, we present a Bayesian approach to estimate these quantities in terms of compositional data.

View in Mathematical Geosciences


a poem, a book, a movie and a song