arXiv subjects

Antonio González

Publications and source records attributed to Antonio González.

27 records · Page 2Linked to original sources

arXiv2021

Exploiting Beam Search Confidence for Energy-Efficient Speech Recognition

With computers getting more and more powerful and integrated in our daily lives, the focus is increasingly shifting towards more human-friendly interfaces, making Automatic Speech Recognition (ASR) a central player as the ideal means of interaction with machines. Consequently, interest in speech technology has grown in the last few years, with more systems being proposed and higher accuracy levels being achieved, even surpassing \textit{Human Accuracy}. While ASR systems become increasingly powerful, the computational complexity also increases, and the hardware support have to keep pace. In this paper, we propose a technique to improve the energy-efficiency and performance of ASR systems, focusing on low-power hardware for edge devices. We focus on optimizing the DNN-based Acoustic Model evaluation, as we have observed it to be the main bottleneck in state-of-the-art ASR systems, by leveraging run-time information from the Beam Search. By doing so, we reduce energy and execution time of the acoustic model evaluation by 25.6% and 25.9%, respectively, with negligible accuracy loss.

cs.SD↗

arXiv2019

Power domination in maximal planar graphs

Power domination in graphs emerged from the problem of monitoring an electrical system by placing as few measurement devices in the system as possible. It corresponds to a variant of domination that includes the possibility of propagation. For measurement devices placed on a set S of vertices of a graph G, the set of monitored vertices is initially the set S together with all its neighbors. Then iteratively, whenever some monitored vertex v has a single neighbor u not yet monitored, u gets monitored. A set S is said to be a power dominating set of the graph G if all vertices of G eventually are monitored. The power domination number of a graph is the minimum size of a power dominating set. In this paper, we prove that any maximal planar graph of order n $\ge$ 6 admits a power dominating set of size at most (n--2)/4 .

cs.DM↗

arXiv2018

Rendering Elimination: Early Discard of Redundant Tiles in the Graphics Pipeline

GPUs are one of the most energy-consuming components for real-time rendering applications, since a large number of fragment shading computations and memory accesses are involved. Main memory bandwidth is especially taxing battery-operated devices such as smartphones. Tile-Based Rendering GPUs divide the screen space into multiple tiles that are independently rendered in on-chip buffers, thus reducing memory bandwidth and energy consumption. We have observed that, in many animated graphics workloads, a large number of screen tiles have the same color across adjacent frames. In this paper, we propose Rendering Elimination (RE), a novel micro-architectural technique that accurately determines if a tile will be identical to the same tile in the preceding frame before rasterization by means of comparing signatures. Since RE identifies redundant tiles early in the graphics pipeline, it completely avoids the computation and memory accesses of the most power consuming stages of the pipeline, which substantially reduces the execution time and the energy consumption of the GPU. For widely used Android applications, we show that RE achieves an average speedup of 1.74x and energy reduction of 43% for the GPU/Memory system, surpassing by far the benefits of Transaction Elimination, a state-of-the-art memory bandwidth reduction technique available in some commercial Tile-Based Rendering GPUs.

cs.AR↗

arXiv2016

Shortcut sets for the locus of plane Euclidean networks

We study the problem of augmenting the locus $\mathcal{N}_{\ell}$ of a plane Euclidean network $\mathcal{N}$ by inserting iteratively a finite set of segments, called \emph{shortcut set}, while reducing the diameter of the locus of the resulting network. There are two main differences with the classical augmentation problems: the endpoints of the segments are allowed to be points of $\mathcal{N}_{\ell}$ as well as points of the previously inserted segments (instead of only vertices of $\mathcal{N}$), and the notion of diameter is adapted to the fact that we deal with $\mathcal{N}_{\ell}$ instead of $\mathcal{N}$. This increases enormously the hardness of the problem but also its possible practical applications to network design. Among other results, we characterize the existence of shortcut sets, compute them in polynomial time, and analyze the role of the convex hull of $\mathcal{N}_{\ell}$ when inserting a shortcut set. Our main results prove that, while the problem of minimizing the size of a shortcut set is NP-hard, one can always determine in polynomial time whether inserting only one segment suffices to reduce the diameter.

math.CO↗

arXiv2016

New results on metric-locating-dominating sets of graphs

A dominating set $S$ of a graph is a metric-locating-dominating set if each vertex of the graph is uniquely distinguished by its distances from the elements of $S$, and the minimum cardinality of such a set is called the metric-location-domination number. In this paper, we undertake a study that, in general graphs and specific families, relates metric-locating-dominating sets to other special sets: resolving sets, dominating sets, locating-dominating sets and doubly resolving sets. We first characterize classes of trees according to certain relationships between their metric-location-domination number and their metric dimension and domination number. Then, we show different methods to transform metric-locating-dominating sets into locating-dominating sets and doubly resolving sets. Our methods produce new bounds on the minimum cardinalities of all those sets, some of them involving parameters that have not been related so far.

math.CO↗

arXiv2014

Resolving sets for breaking symmetries of graphs

This paper deals with the maximum value of the difference between the determining number and the metric dimension of a graph as a function of its order. Our technique requires to use locating-dominating sets, and perform an independent study on other functions related to these sets. Thus, we obtain lower and upper bounds on all these functions by means of very diverse tools. Among them are some adequate constructions of graphs, a variant of a classical result in graph domination and a polynomial time algorithm that produces both distinguishing sets and determining sets. Further, we consider specific families of graphs where the restrictions of these functions can be computed. To this end, we utilize two well-known objects in graph theory: $k$-dominating sets and matchings.

math.CO↗

arXiv2013

The resolving number of a graph

We study a graph parameter related to resolving sets and metric dimension, namely the resolving number, introduced by Chartrand, Poisson and Zhang. First, we establish an important difference between the two parameters: while computing the metric dimension of an arbitrary graph is known to be NP-hard, we show that the resolving number can be computed in polynomial time. We then relate the resolving number to classical graph parameters: diameter, girth, clique number, order and maximum degree. With these relations in hand, we characterize the graphs with resolving number 3 extending other studies that provide characterizations for smaller resolving number.

math.CO↗

arXiv2012

Resolving sets for Johnson and Kneser graphs

A set of vertices $S$ in a graph $G$ is a {\em resolving set} for $G$ if, for any two vertices $u,v$, there exists $x\in S$ such that the distances $d(u,x) \neq d(v,x)$. In this paper, we consider the Johnson graphs $J(n,k)$ and Kneser graphs $K(n,k)$, and obtain various constructions of resolving sets for these graphs. As well as general constructions, we show that various interesting combinatorial objects can be used to obtain resolving sets in these graphs, including (for Johnson graphs) projective planes and symmetric designs, as well as (for Kneser graphs) partial geometries, Hadamard matrices, Steiner systems and toroidal grids.

math.CO↗

arXiv2012

On the metric dimension, the upper dimension and the resolving number of graphs

This paper deals with three resolving parameters: the metric dimension, the upper dimension and the resolving number. We first answer a question raised by Chartrand and Zhang asking for a characterization of the graphs with equal metric dimension and resolving number. We also solve in the affirmative a conjecture posed by Chartrand, Poisson and Zhang about the realization of the metric dimension and the upper dimension. Finally we prove that no integer $a\geq 4$ is realizable as the resolving number of an infinite family of graphs.

math.CO↗