arXiv subjects

Megha Khosla

Publications and source records attributed to Megha Khosla.

41 records · Page 3Linked to original sources

arXiv2019

Node Representation Learning for Directed Graphs

We propose a novel approach for learning node representations in directed graphs, which maintains separate views or embedding spaces for the two distinct node roles induced by the directionality of the edges. We argue that the previous approaches either fail to encode the edge directionality or their encodings cannot be generalized across tasks. With our simple \emph{alternating random walk} strategy, we generate role specific vertex neighborhoods and train node embeddings in their corresponding source/target roles while fully exploiting the semantics of directed graphs. We also unearth the limitations of evaluations on directed graphs in previous works and propose a clear strategy for evaluating link prediction and graph reconstruction in directed graphs. We conduct extensive experiments to showcase our effectiveness on several real-world datasets on link prediction, node classification and graph reconstruction tasks. We show that the embeddings from our approach are indeed robust, generalizable and well performing across multiple kinds of tasks and graphs. We show that we consistently outperform all baselines for node classification task. In addition to providing a theoretical interpretation of our method we also show that we are considerably more robust than the other directed graph approaches.

cs.SI↗

arXiv2018

Asynchronous Training of Word Embeddings for Large Text Corpora

Word embeddings are a powerful approach for analyzing language and have been widely popular in numerous tasks in information retrieval and text mining. Training embeddings over huge corpora is computationally expensive because the input is typically sequentially processed and parameters are synchronously updated. Distributed architectures for asynchronous training that have been proposed either focus on scaling vocabulary sizes and dimensionality or suffer from expensive synchronization latencies. In this paper, we propose a scalable approach to train word embeddings by partitioning the input space instead in order to scale to massive text corpora while not sacrificing the performance of the embeddings. Our training procedure does not involve any parameter synchronization except a final sub-model merge phase that typically executes in a few minutes. Our distributed training scales seamlessly to large corpus sizes and we get comparable and sometimes even up to 45% performance improvement in a variety of NLP benchmarks using models trained by our distributed procedure which requires $1/10$ of the time taken by the baseline approach. Finally we also show that we are robust to missing words in sub-models and are able to effectively reconstruct word representations.

cs.LG↗

arXiv2018

User Fairness in Recommender Systems

Recent works in recommendation systems have focused on diversity in recommendations as an important aspect of recommendation quality. In this work we argue that the post-processing algorithms aimed at only improving diversity among recommendations lead to discrimination among the users. We introduce the notion of user fairness which has been overlooked in literature so far and propose measures to quantify it. Our experiments on two diversification algorithms show that an increase in aggregate diversity results in increased disparity among the users.

cs.CY↗

arXiv2014

Macroscopic quantum information processing using spin coherent states

Previously a new scheme of quantum information processing based on spin coherent states of two component Bose-Einstein condensates was proposed (Byrnes {\it et al.} Phys. Rev. A 85, 40306(R)). In this paper we give a more detailed exposition of the scheme, expanding on several aspects that were not discussed in full previously. The basic concept of the scheme is that spin coherent states are used instead of qubits to encode qubit information, and manipulated using collective spin operators. The scheme goes beyond the continuous variable regime such that the full space of the Bloch sphere is used. We construct a general framework for quantum algorithms to be executed using multiple spin coherent states, which are individually controlled. We illustrate the scheme by applications to quantum information protocols, and discuss possible experimental implementations. Decoherence effects are analyzed under both general conditions and for the experimental implementation proposed.

quant-ph↗

arXiv2013

The Multiple-orientability Thresholds for Random Hypergraphs

A $k$-uniform hypergraph $H = (V, E)$ is called $\ell$-orientable, if there is an assignment of each edge $e\in E$ to one of its vertices $v\in e$ such that no vertex is assigned more than $\ell$ edges. Let $H_{n,m,k}$ be a hypergraph, drawn uniformly at random from the set of all $k$-uniform hypergraphs with $n$ vertices and $m$ edges. In this paper we establish the threshold for the $\ell$-orientability of $H_{n,m,k}$ for all $k\ge 3$ and $\ell \ge 2$, i.e., we determine a critical quantity $c_{k, \ell}^*$ such that with probability $1-o(1)$ the graph $H_{n,cn,k}$ has an $\ell$-orientation if $c < c_{k, \ell}^*$, but fails doing so if $c > c_{k, \ell}^*$. Our result has various applications including sharp load thresholds for cuckoo hashing, load balancing with guaranteed maximum load, and massive parallel access to hard disk arrays.

cs.DM↗