Hi, I'm Miguel Romero!
I'm a PhD student at the Université Côte d'Azur in France since 2017, I work under the supervision of Professors Frédéric Precioso, Lucile Sassatelli and Ramón Aparicio. My project is in the field of Machine Learning and Network Optimization, specifically in the use of Deep Learning architectures to optimize the streaming of Virtual Reality videos.
About me
I am a Computer Systems Engineer since 2014; from 2012 to 2014 I worked on Tele-Medicine projects involving panoramic image generation, real-time feature detection, medical image processing  and robotics in the Research Group in Biomedical Engineering of the Industrial University of Santander, Colombia.

In 2015 I received the LabEx scholarship to study at the University of Nice-Sophia Antipolis, where I worked on the project "Green Optimal Baseband Unit Placement for Fixed/Mobile Converged Aggregation Networks" to optimize the video streaming delivery by placing the computation units in the network in a way that optimizes the energy consumption and the Quality of Experience of the viewer.

During the summer of 2016, I did an internship to work on the problem of "Interactive Content-based Retrieval from Eye-tracking". For the internship I applied Active Learning to refine the selection of displayed images and make the system converging more rapidly to the target images, using an eye tracker to follow the gaze of the user.

In 2017, for my second year of the International Master in Ubiquitous Networking. I worked on the project "Learning from Experiments on Machine Learning Workflows" as a contribution to the LabEx project RockFlows, to create a Meta-Model that proposes the best Machine Learning Workflow given an input dataset.

I started my PhD on November 2017, my thesis "Streaming Virtual Reality: Learning for Attentional Models and Network Optimization" I research ways to structure VR content in the cloud and optimally decide what to transmit to maximize the streaming quality. Our strategy consists in devising a comprehensive processing chain tailored to VR and targeted at streaming, including concepts like Foveated Streaming, Gaze Guidance and Head Motion prediction of viewers in 360° videos using Deep Learning.
Streaming Virtual Reality
Foveated Streaming
Learning for Attentional Models and Network Optimization.
Meta-Learning
Learning from ML Experiments
PFE: Learning from experiments on machine learning workflows, towards a software product line for ML workflows.