Author name cluster

Harsh H. Pareek

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

3 papers

1 author row

ICML Conference 2015 Conference Paper

Distributional Rank Aggregation, and an Axiomatic Analysis

Adarsh Prasad
Harsh H. Pareek
Pradeep Ravikumar

The rank aggregation problem has been studied with varying desiderata in varied communities such as Theoretical Computer Science, Statistics, Information Retrieval and Social Welfare Theory. We introduce a variant of this problem we call distributional rank aggregation, where the ranking data is only available via the induced distribution over the set of all permutations. We provide a novel translation of the usual social welfare theory axioms to this setting. As we show this allows for a more quantitative characterization of these axioms: which then are not only less prone to misinterpretation, but also allow simpler proofs for some key impossibility theorems. Most importantly, these quantitative characterizations lead to natural and novel relaxations of these axioms, which as we show, allow us to get around celebrated impossibility results in social choice theory. We are able to completely characterize the class of positional scoring rules with respect to our axioms and show that Borda Count is optimal in a certain sense.

Details

UAI Conference 2015 Conference Paper

Tracking with ranked signals

Tianyang Li
Harsh H. Pareek
Pradeep Ravikumar
Dhruv Balwada
Kevin Speer

We present a novel graphical model approach for a problem not previously considered in the machine learning literature: that of tracking with ranked signals. The problem consists of tracking a single target given observations about the target that consist of ranked continuous signals, from unlabeled sources in a cluttered environment. We introduce appropriate factors to handle the imposed ordering assumption, and also incorporate various systematic errors that can arise in this problem, particularly clutter or noise signals as well as missing signals. We show that inference in the obtained graphical model can be simplified by adding bipartite structures with appropriate factors. We apply a hybrid approach consisting of belief propagation and particle filtering in this mixed graphical model for inference and validate the approach on simulated data. We were motivated to formalize and study this problem by a key task in Oceanography, that of tracking the motion of RAFOS ocean floats, using range measurements sent from a set of fixed beacons, but where the identities of the beacons corresponding to the measurements are not known. However, unlike the usual tracking problem in artificial intelligence, there is an implicit ranking assumption among signal arrival times. Our experiments show that the proposed graphical model approach allows us to effectively leverage the problem constraints and improve tracking accuracy over baseline tracking methods yielding results similar to the ground truth hand-labeled data.

Details

ICML Conference 2013 Conference Paper

Human Boosting

Harsh H. Pareek
Pradeep Ravikumar

Humans may be exceptional learners but they have biological limitations and moreover, inductive biases similar to machine learning algorithms. This puts limits on human learning ability and on the kinds of learning tasks humans can easily handle. In this paper, we consider the problem of “boosting” human learners to extend the learning ability of human learners and achieve improved performance on tasks which individual humans find difficult. We consider classification (category learning) tasks, propose a boosting algorithm for human learners and give theoretical justifications. We conduct experiments using Amazon’s Mechanical Turk on two synthetic datasets – a crosshair task with a nonlinear decision boundary and a gabor patch task with a linear boundary but which is inaccessible to human learners – and one real world dataset – the Opinion Spam detection task introduced in (Ott et al). Our results show that boosting human learners produces gains in accuracy and can overcome some fundamental limitations of human learners.

Details