Rishabh Iyer

University of Washington, Machine Learning, Post-Doc

Followers

Following

Public Views

I recently finished my Ph.D at the University of Washington, Seattle, where I worked with Prof. Jeff Bilmes. My main interests are at the intersection of Machine Learning and submodular and discrete optimization. For some information, feel free to visit https://sites.google.com/view/rishabhiyer/home
Supervisors: Prof. Jeff Bilmes

less

Judith L Green

University of California, Santa Barbara

Steven Pinker

Harvard University

E. Wayne Ross

University of British Columbia

Andreas Krause

Swiss Federal Institute of Technology (ETH)

Armando Marques-Guedes

UNL - New University of Lisbon

Fabio Cuzzolin

Oxford Brookes University

Roshan Chitrakar

Nepal College of Information Technology

Lev Manovich

Graduate Center of the City University of New York

Soumyajit De

IIT Bombay

Bogdan Gabrys

University of Technology Sydney

InterestsView All (6)

Uploads

Papers by Rishabh Iyer

Submodular-Bregman and the Lov sz-Bregman

Submodular-Bregman and the Lov´ sz-Bregman a Divergences with Applications Jeff Bilmes Department... more Submodular-Bregman and the Lov´ sz-Bregman a Divergences with Applications Jeff Bilmes Department of Electrical Engineering University of Washington [email protected] Rishabh Iyer Department of Electrical Engineering University of Washington [email protected] Abstract We introduce a class of discrete divergences on sets (equivalently binary vectors) that we call the submodular-Bregman divergences.

Object Mining for Large Video Data

The task of object mining has attracted a lot of attention of vision researchers. An efficient ob... more The task of object mining has attracted a lot of attention of vision researchers. An efficient object mining scheme has many applications, including a mehanism for accessing data in an organized manner, selective video playback and summarization. Traditionally this task has been attempted in two ways. Several approaches cast this problem as one of extracting frequently occurring patterns. The over-all problem is then cast in the manner analogous to searching in a text corpus [7],[8],[5].

Download

Estimation of the Embedding Capacity in Pixel-pair based Watermarking Schemes

Abstract: Estimation of the Embedding capacity is an important problem specifically in reversible... more Abstract: Estimation of the Embedding capacity is an important problem specifically in reversible multi-pass watermarking and is required for analysis before any image can be watermarked. In this paper, we propose an efficient method for estimating the embedding capacity of a given cover image under multi-pass embedding, without actually embedding the watermark. We demonstrate this for a class of reversible watermarking schemes which operate on a disjoint group of pixels, specifically for pixel pairs.

Mirror Descent-Like Algorithms for Submodular Optimization

In this paper we develop a framework of submodular optimization algorithms in line with the mirro... more In this paper we develop a framework of submodular optimization algorithms in line with the mirror-descent style of algorithms for convex optimization. We use the fact that a submodular function has both a subdifferential and a superdifferential, which enables us to formulate algorithms for both submodular minimization and maximization. This reveals a unifying framework for a number of submodular maximization algorithms. In addition, we refine a number of bounds for constrained submodular minimization.

Download

Submodular-Bregman and the Lovasz-Bregman Divergences with Applications

We introduce a class of discrete divergences on sets (equivalently binary vectors) that we call t... more We introduce a class of discrete divergences on sets (equivalently binary vectors) that we call the submodular-Bregman divergences. We consider two kinds, defined either from tight modular upper or tight modular lower bounds of a submodular function. We show that the properties of these divergences are analogous to the (standard continuous) Bregman divergence. We demonstrate how they generalize many useful divergences, including the weighted Hamming distance, squared weighted Hamming, weighted precision, recall, conditional mutual information, and a generalized KL-divergence on sets. We also show that the generalized Bregman divergence on the Lovász extension of a submodular function, which we call the Lovász-Bregman divergence, is a continuous extension of a submodular Bregman divergence. We point out a number of applications, and in particular show that a proximal algorithm defined through the submodular Bregman divergence provides a framework for many mirror-descent style algorithms related to submodular function optimization. We also show that a generalization of the k-means algorithm using the Lovász Bregman divergence is natural in clustering scenarios where ordering is important. A unique property of this algorithm is that computing the mean ordering is extremely efficient unlike other order based distance measures.

Download

Algorithms for Approximate Minimization of the Difference Between Submodular Functions, with Applications

We extend the work of Narasimhan and Bilmes for minimizing set functions representable as a diffe... more We extend the work of Narasimhan and Bilmes for minimizing set functions representable as a difference between submodular functions. Similar to , our new algorithms are guaranteed to monotonically reduce the objective function at every step. We empirically and theoretically show that the per-iteration cost of our algorithms is much less than [30], and our algorithms can be used to efficiently minimize a difference between submodular functions under various combinatorial constraints, a problem not previously addressed. We provide computational bounds and a hardness result on the multiplicative inapproximability of minimizing the difference between submodular functions. We show, however, that it is possible to give worst-case additive bounds by providing a polynomial time computable lower-bound on the minima. Finally we show how a number of machine learning problems can be modeled as minimizing the difference between submodular functions. We experimentally show the validity of our algorithms by testing them on the problem of feature selection with submodular cost features.

Download