Skip to content

Esteban Bautista

Researcher in Data Science

I am a postdoctoral researcher in the DECIDE team at IMT Atlantique (France).

My research focuses on the learning, modeling, and processing of data having both temporal and relational components, such as network traffic, phone calls, financial transactions, etc. My goals are (i) to develop adapted tools for the study of such data; (ii) to develop explainable learning methods that require minimum human intervention; (iii) to develop algorithms that scale to massive datasets; and (iv) to detect important events in such data.

Interests

  • Link Streams
  • Machine Learning 
  • Signal Processing
  • Explainable AI
  • Graph Theory and Algorithms
  • Network Science
  • Anomaly Detection

Education

ENS de Lyon, under Paulo Gonçalves and Patrice Abry

National Autonomous University of Mexico (UNAM)

National Autonomous University of Mexico (UNAM)

News

I am happy to announce that our paper

E. Bautista. L. Brisson, C. Bothorel, G. Smits, “MAD: Multi-Scale Anomaly Detection in Link Streams” (link)

has been accepted at The 17th ACM International Conference on Web Search and Data Mining (WSDM). 

I am happy to announce that our paper

N. Arhachoui, E. Bautista, M. Danisch, A. Giovanidis, L. Tabourier, “Scalable Algorithms to Measure User Influence in Social Networks

has been accepted at Lecture Notes in Social Networks, Springer. 

I am happy to announce that our paper

E. Bautista, M. Latapy, “Link Streams as a Generalization of Graphs and Time Series” (link)

has been accepted at the The Fifth IEEE International Conference on Cognitive Machine Intelligence (COGMI) 2023. 

I am happy to announce that our paper

E. Bautista, M. Latapy, “A Frequency-Structure Approach for Link Stream Analysis” (link)

has been accepted at the 2nd edition of Temporal Network Theory, Springer book.

I am happy to announce that I will be joining the Decide team of the Lab-STICC at IMT Atlantique as a postdoctoral researcher. 

I will be working on developing new algorithms for detecting anomalous events in interaction streams (temporal networks). 

Looking forward for an exciting collaboration with Cécile Bothorel, Laurent Brisson and Grégory Smits.

I am happy to announce that our paper

N. Arhachoui, E. Bautista, M. Danisch. A. Giovanidis, “A Fast Algorithm for Ranking Users by Their Influence in Online Social Platforms” (link)

has been accepted at the 2022 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM).

I invite everyone to the seminar I will give at the CNA discussion group of the Demokritos Research Center, Greece. 

Title: A frequency-structure decomposition for link streams

Abstract: A link stream is a set of triplets (t, u, v) modeling interactions over time, such as person u calling v at time t, or bank account u transferring to v at time t. Effectively analyzing link streams is thus key for numerous applications. In practice, it is common to study link streams as a collection of time series or as a sequence of graphs, allowing to use time filters and graph filters to process the time and structural dimensions, respectively. However, time and structure are nested in link streams, meaning that time-domain operations can affect structure, and vice-versa. This calls for a frequency-structure representation that allows to characterize processing operations in both frequency and structure. Yet, it is hard to combine existing signal and graph decompositions as they do not interact well. 

To address this limitation, this work proposes a novel frequency-structure decomposition for link streams. Our decomposition allows us to analyze time via existing signal decompositions (Fourier, Wavelets, etc) and to analyze structure via a novel decomposition for graphs that is tailor-made to interact well with signal decompositions. This novel graph decomposition operates by partitioning the edge-space of graphs into regions and measuring the activity of regions, resulting in a set of coefficients that have several interesting properties to characterize the structural properties of graphs and that can be used to compare them. We show that the combination of our graph decomposition with signal decompositions leads to a set of coefficients that effectively encode the dynamical and structural properties of link streams in a simple matrix format. Then, we show that the properties of our decomposition make it easy to define time and graph filters in the frequency-structure domain and finish by showing how we can use filters to recover the backbone of a link stream.

I am happy to announce that our communication

E. Bautista, M. Latapy, “A Frequency-Structure decomposition For Link Streams” (link)

has been accepted as an oral presentation at the colloquium GRETSI 2022.

I invite everyone to the seminar I will give at Université de Bourgogne / LIB Laboratory.

Title: A frequency-structure decomposition for link streams

Abstract: A link stream is a set of triplets (t, u, v) modeling interactions over time, such as person u calling v at time t, or bank account u transferring to v at time t. Effectively analyzing link streams is thus key for numerous applications. In practice, it is common to study link streams as a collection of time series or as a sequence of graphs, allowing to use time filters and graph filters to process the time and structural dimensions, respectively. However, time and structure are nested in link streams, meaning that time-domain operations can affect structure, and vice-versa. This calls for a frequency-structure representation that allows to characterize processing operations in both frequency and structure. Yet, it is hard to combine existing signal and graph decompositions as they do not interact well. 

To address this limitation, this work proposes a novel frequency-structure decomposition for link streams. Our decomposition allows us to analyze time via existing signal decompositions (Fourier, Wavelets, etc) and to analyze structure via a novel decomposition for graphs that is tailor-made to interact well with signal decompositions. This novel graph decomposition operates by partitioning the edge-space of graphs into regions and measuring the activity of regions, resulting in a set of coefficients that have several interesting properties to characterize the structural properties of graphs and that can be used to compare them. We show that the combination of our graph decomposition with signal decompositions leads to a set of coefficients that effectively encode the dynamical and structural properties of link streams in a simple matrix format. Then, we show that the properties of our decomposition make it easy to define time and graph filters in the frequency-structure domain and finish by showing how we can use filters to recover the backbone of a link stream.

 

I am happy to announce that our paper

E. Bautista, M. Latapy, ‘A Local Updating Algorithm for Personalized PageRank via Chebyshev Polynomials‘ (link)

has been accepted for publication at Social Network Analysis and Mining Journal, Springer.