^May June – 2012 ^July

Our visual attention is attracted by salient stimuli in our environment and affected by primitive features such as orientation, color, and motion. Perceptual saliency due to orientation contrast has been extensively demonstrated in behavioral experiments with humans and other primates and is commonly explained by the very particular functional organization of the primary visual cortex. We challenge this prevailing view by studying orientation-based visual saliency in two non-mammalian species with enormous evolutionary distance to humans. The surprising results not only imply the need to reestablish our understanding of how these processes work at the neural level, but they also suggest that orientation-based saliency has computational optimality in a wide variety of ecological contexts, and thus constitutes a universal building block for efficient visual information processing in general.

We show that an ``old dog'', namely – the classical Johnson-Lindenstrauss transform, ``performs new tricks'' – it gives a novel way of preserving differential privacy. We show that if we take two databases, D and D', such that (i) D'-D is a rank-1 matrix of bounded norm and (ii) all singular values of D and D' are sufficiently large, then multiplying either D or D' with a vector of iid normal Gaussians yields two statistically close distributions in the sense of differential privacy. Furthermore, a small, deterministic and public alteration of the input is enough to assert that all singular values of D are large. We apply the Johnson-Lindenstrauss transform to the task of approximating cut-queries: the number of edges crossing a (S,bar S)-cut in a graph. We show that the JL transform allows us to publish a sanitized graph that preserves edge differential privacy (where two graphs are neighbors if they differ on a single edge) while adding only O(|S|/epsilon) random noise to any given query (w.h.p). Comparing the additive noise of our algorithm to existing algorithms for answering cut-queries in a differentially private manner, we outperform all others on small cuts (|S| = o(n)). We also apply our technique to the task of estimating the variance of a given matrix in any given direction. The JL transform allows us to publish a sanitized covariance matrix that preserves differential privacy w.r.t bounded changes (each row in the matrix can change by at most a norm-1 vector) while adding random noise of magnitude independent of the size of the matrix (w.h.p). In contrast, existing algorithms introduce an error which depends on the matrix dimensions.

Despite their gross simplicity, random (linear) codes gained great fame with their use in Shannon's 1948 noisy-channel coding theorem. In this talk, we survey three of our applications of these codes to the field of networks. Our first work is a concrete implementation of random coding for over-the-air programming of sensor motes. Our second work relates to fair and secure bandwidth sharing of assymetric channels, where we utilize a game-theoretic framework to craft a protocol resistant to maliciously colluding parties. Our third application involves the use of extreme value theory to predict system-level error rates for wireless broadcast. Each case serves to demonstrate the astonishing power that can be harnessed from these amazingly simple codes.

Inferring drug-associated traits such as drug targets, drug indications and drug-drug interactions is essential for drug development and drug administration. I will present a novel method for the large-scale prediction of such traits able to handle both approved drugs and novel molecules. Our method is based on the observation that similar drugs tend to associate with similar traits and utilizes various similarity measures for the prediction task. Furthermore, our method is utilized for inferring causative factors of drug-drug interactions and can be extended to handle personalized medicine.