Sorry, you need to enable JavaScript to visit this website.

Differential privacy is a strong, cryptographically-motivated definition of privacy that has recently received a significant amount of research attention for its robustness to known attacks. The principal component analysis (PCA) algorithm is frequently used in signal processing, machine learning and statistics pipelines. In this paper, we propose a new algorithm for differentially-private computation of PCA and compare the performance empirically with some recent state-of-the-art algorithms on different data sets.


Recent work on zero resource word discovery makes intensive use of audio fragment clustering to find repeating speech patterns. In the absence of acoustic models, the clustering step traditionally relies on dynamic time warping (DTW) to compare two samples and thus suffers from the known limitations of this technique. We propose a new sample comparison method, called 'similarity by terative classification', that exploits the modeling capacities of hidden Markov models (HMM) with no supervision.


In this paper, we present Discriminant Correlation Analysis (DCA), a feature level fusion technique that incorporates the class associations in correlation analysis of the feature sets. DCA performs an effective feature fusion by maximizing the pair-wise correlations across the two feature sets, and at the same time, eliminating the between-class correlations and restricting the correlations to be within classes.


Imaging techniques involve counting of photons striking a detector.
Due to fluctuations in the counting process, the measured
photon counts are known to be corrupted by Poisson
noise. In this paper, we propose a blind dictionary learning
framework for the reconstruction of photographic image data
from Poisson corrupted measurements acquired by a compressive
camera. We exploit the inherent non-negativity of the
data by modeling the dictionary as well as the sparse dictionary
coefficients as non-negative entities, and infer these directly


Plug-meters benefit many grid and building-level energy management applications like automated load control and load scheduling. However, installing and maintaining large and/orlong term deployments of such meters requires assignment and updating of the identity (labels) of electrical loads connected to them. Although the literature on electricity disaggregation and appliance identification is extensive, there is no consensus on the generalizability of the proposed solutions, especially with respect to the features that are extracted from voltage and current measurements.


Greedy pursuit is one of the standard approaches for sparse approximation. Since the derivation of the conventional greedy pursuit schemes, including matching pursuit (MP) and orthogonal MP (OMP), is based on the inner product space, they may not work properly in the presence of impulsive noise. In this work, we devise a new definition of correlation in lp-space with p>0, called lp-correlation, and introduce the concept of orthogonality in lp-space.


Presentation slides covering:

- robust foreground detection / background subtraction via patch-based analysis
- person re-identification based on representations on Riemannian manifolds
- robust object tracking via Grassmann manifolds
- adapting the lessons from big data to computer vision
- future paradigm shifts: computer vision based on networks of neurosynaptic cores


Slides from the Tutorial on Riemannian Geometry in Computer Vision, presented at the Asian Conference on Computer Vision (ACCV), Singapore, 2014.

The slides show (1) how objects can be interpreted as points on Riemannian and Grassmann manifolds, and (2) various distance measures on manifolds. Demonstrates usefulness of manifold techniques in applications such as object tracking and person re-identification.