Research

My research projects: code, data & papers

This page contains my research projects, my papers (together with the code to reproduce results and figures), slides of my talks and the PDFs of the posters I present at conferences. The material is organized by research topic.

Sparse phase retrieval (2011-still ongoing)

Joint work with M. Vetterli, A. Chebira

TO BE COMPLETED

Euclidean distance matrices

Joint work with I. Dokmanić (main author), R. Parhizkar, M. Vetterli

We have written a review paper on EDM applied to signal processing problems on IEEE Signal Processing Magazine. You can find it here. We are going to release soon more material about this, including many Iptyhon notebooks implementing the described algorithms.

Sensor placement (2011-2014)

Joint work with M. Vetterli, D. Atienza, A. Chebira, Z. Chen, R. Zhang, A. Vincenzi

Assume we deploy a sensor network and we can choose the position of the nodes from a set of possible locations. We would like to optimize the sensor locations such that we minimize the reconstruction error when solving a given inverse problem. Choosing the sensor locations under such constraints is a combinatorial problem and only a brute force approach can find the optimal solution in polynomial time.

Our contributions are:

  1. A near-optimal algorithm for the sensor placement for linear inverse problems. We discuss our result in this paper and the code to reproduce the figures of the paper is available here. In sum, the paper introduces the prolem, presents a greedy algorithm based on the frame potential (a cost function used in frame theory) and proves that such an algorithm is theoretically near-optimal. We corroborate our results with numerical experiments on random and real-world linear inverse problems. An extension of this work to union-of-subspaces has been presented at EUSIPCO 2014.

  2. A state-of-the-art method for the thermal monitoring of many-core microprocessors. We discuss our results on two main publications: one at DAC 2012 and one on IEEE Transactions on Computers. In these papers, we describe methods and algorithms to optimize the sensor placement

  3. A state-of-the-art method for the adaptive sampling scheduling of sensor network.

Inverse problems of the diffusion equation (2009-2013)

Joint work with M. Vetterli, A. Chebira, Y. M. Lu, I. Dokmanić

Assume we deploy a sensor network that collects measurements about a diffusive process induced by unknown sources.

comments powered by Disqus