My research interests include signal processing, bioinformatics, machine learning, and communications. Recent research thrusts include:
More details can be found in Projects and Publications.
Base-calling algorithms for next-generation DNA sequencing systems.
Novel sequencing technologies will provide significant improvements in many aspects of human condition, ultimately
leading towards the understanding, diagnosis, treatment, and prevention of diseases. Reliable decision-making in
such downstream applications is predicated upon accurate base-calling, i.e., precise identification of the order of
nucleotides from noisy sequencing data. We explored trade-offs between performance and complexity of base-calling,
developing algorithms that have both better accuracy and
are orders of magnitude faster than
- Signal processing for real-time biosensor arrays.
Real-time DNA microarrays observe
the dynamics of binding between molecular targets and the biosensing elements that capture them. The paradigm shift
in data acquisition, from measuring a single steady-state data point in conventional systems to temporally sampling
dynamics of the binding process, requires
new mathematical models and inference methods. We developed fast algorithms
that enable reliable estimation of the number of molecular targets in real-time biosensor arrays, demonstrating
significant performance gains over the conventional technology.
- Sensor and cognitive radio networks. For a rational use of the sensor network resources,
the fusion center typically schedules only a subset of the available sensors for transmission, and the selected sensors
transmit only partial information. Recently, we showed that in a variety of scenarios sensor selection can be cast as
maximization of a submodular function over uniform matroids. It turns out that a computationally efficient greedy sensor
selection algorithm is guaranteed to perform within (1-1/e) of the optimal solution. We extended these results to the
scenario where the communication between sensors and the fusion center is unreliable.
- Algorithms for solving integer least-squares problems in communications.
Complexity of the optimal signal processing in communication systems is a main obstacle to delivering the high data rates
promised by theory. In a probabilistic setting, common to communications, it is possible to exploit statistics of the problem
in order to design fast algorithms and analyze their complexity. An example of our work in this area includes the analysis of
the expected complexity of an algorithm for optimal decoding in high-dimensional communication systems.
The results reveal that
optimal decoding, hitherto considered infeasible, can in fact often be implemented in practice.