arXiv subjects

Antonio Fernandez Anta

Publications and source records attributed to Antonio Fernandez Anta.

15 recordsLinked to original sources

arXiv2024

Federated Learning with Heterogeneous Data Handling for Robust Vehicular Object Detection

In the pursuit of refining precise perception models for fully autonomous driving, continual online model training becomes essential. Federated Learning (FL) within vehicular networks offers an efficient mechanism for model training while preserving raw sensory data integrity. Yet, FL struggles with non-identically distributed data (e.g., quantity skew), leading to suboptimal convergence rates during model training. In previous work, we introduced FedLA, an innovative Label-Aware aggregation method addressing data heterogeneity in FL for generic scenarios. In this paper, we introduce FedProx+LA, a novel FL method building upon the state-of-the-art FedProx and FedLA to tackle data heterogeneity, which is specifically tailored for vehicular networks. We evaluate the efficacy of FedProx+LA in continuous online object detection model training. Through a comparative analysis against conventional and state-of-the-art methods, our findings reveal the superior convergence rate of FedProx+LA. Notably, if the label distribution is very heterogeneous, our FedProx+LA approach shows substantial improvements in detection performance compared to baseline methods, also outperforming our previous FedLA approach. Moreover, both FedLA and FedProx+LA increase convergence speed by 30% compared to baseline methods.

cs.CV↗

arXiv2021

MonTrees: Automated Detection and Classification of Networking Anomalies in Cellular Networks

The active growth and dynamic nature of cellular networks makes network troubleshooting challenging. Identification of network problems leveraging on machine learning has gained a lot of visibility in the past few years, resulting in dramatically improved cellular network services. In this paper, we present a novel methodology to automate the fault identification process in a cellular network and to classify network anomalies, which combines supervised and unsupervised machine learning algorithms. Our experiments using real data from operational commercial mobile networks obtained through drive-test measurements as well as via the MONROE platform show that our method can automatically identify and classify networking anomalies, thus enabling timely and precise troubleshooting actions.

cs.NI↗

arXiv2021

Fragmented Objects: Boosting Concurrency of Shared Large Objects

This work examines strategies to handle large shared data objects in distributed storage systems (DSS), while boosting the number of concurrent accesses, maintaining strong consistency guarantees, and ensuring good operation performance. To this respect, we define the notion of fragmented objects:con-current objects composed of a list of fragments (or blocks) that allow operations to manipulate each of their fragments individually. As the fragments belong to the same object, it is not enough that each fragment is linearizable to have useful consistency guarantees in the composed object. Hence, we capture the consistency semantic of the whole object with the notion of fragmented linearizability. Then, considering that a variance of linearizability, coverability, is more suited for versioned objects like files, we provide an implementation of a distributed file system, called COBFS, that utilizes coverable fragmented objects (i.e., files).In COBFS, each file is a linked-list of coverable block objects. Preliminary emulation of COBFS demonstrates the potential of our approach in boosting the concurrency of strongly consistent large objects.

cs.DC↗

arXiv2020

Appending Atomically in Byzantine Distributed Ledgers

A Distributed Ledger Object (DLO) is a concurrent object that maintains a totally ordered sequence of records, and supports two basic operations: append, which appends a record at the end of the sequence, and get, which returns the sequence of records. In this work we provide a proper formalization of a Byzantine-tolerant Distributed Ledger Object (BDLO), which is a DLO in a distributed system in which processes may deviate arbitrarily from their indented behavior, i.e. they may be Byzantine. Our formal definition is accompanied by algorithms to implement BDLOs by utilizing an underlying Byzantine Atomic Broadcast service. We then utilize the BDLO implementations to solve the Atomic Appends problem against Byzantine processes. The Atomic Appends problem emerges when several clients have records to append, the record of each client has to be appended to a different BDLO, and it must be guaranteed that either all records are appended or none. We present distributed algorithms implementing solutions for the Atomic Appends problem when the clients (which are involved in the appends) and the servers (which maintain the BDLOs) may be Byzantine.

cs.DC↗

arXiv2019

A Novel Hyperparameter-free Approach to Decision Tree Construction that Avoids Overfitting by Design

Decision trees are an extremely popular machine learning technique. Unfortunately, overfitting in decision trees still remains an open issue that sometimes prevents achieving good performance. In this work, we present a novel approach for the construction of decision trees that avoids the overfitting by design, without losing accuracy. A distinctive feature of our algorithm is that it requires neither the optimization of any hyperparameters, nor the use of regularization techniques, thus significantly reducing the decision tree training time. Moreover, our algorithm produces much smaller and shallower trees than traditional algorithms, facilitating the interpretability of the resulting models.

cs.LG↗

arXiv2018

Atomic Appends: Selling Cars and Coordinating Armies with Multiple Distributed Ledgers

The various applications using Distributed Ledger Technologies (DLT) or blockchains, have led to the introduction of a new `marketplace' where multiple types of digital assets may be exchanged. As each blockchain is designed to support specific types of assets and transactions, and no blockchain will prevail, the need to perform interblockchain transactions is already pressing. In this work we examine the fundamental problem of interoperable and interconnected blockchains. In particular, we begin by introducing the Multi-Distributed Ledger Objects (MDLO), which is the result of aggregating multiple Distributed Ledger Objects -- DLO (a DLO is a formalization of the blockchain) and that supports append and get operations of records (e.g., transactions) in them from multiple clients concurrently. Next, we define the AtomicAppends problem, which emerges when the exchange of digital assets between multiple clients may involve appending records in more than one DLO. Specifically, AtomicAppend requires that either all records will be appended on the involved DLOs or none. We examine the solvability of this problem assuming rational and risk-averse clients that may fail by crashing, and under different client utility and append models, timing models, and client failure scenarios. We show that for some cases the existence of an intermediary is necessary for the problem solution. We propose the implementation of such intermediary over a specialized blockchain, we term Smart DLO (SDLO), and we show how this can be used to solve the AtomicAppends problem even in an asynchronous, client competitive environment, where all the clients may crash.

cs.DC↗

arXiv2018

Coping with Unreliable Workers in Internet-based Computing: An Evaluation of Reputation Mechanisms

We present reputation-based mechanisms for building reliable task computing systems over the Internet. The most characteristic examples of such systems are the volunteer computing and the crowdsourcing platforms. In both examples end users are offering over the Internet their computing power or their human intelligence to solve tasks either voluntarily or under payment. While the main advantage of these systems is the inexpensive computational power provided, the main drawback is the untrustworthy nature of the end users. Generally, this type of systems are modeled under the "master-worker" setting. A "master" has a set of tasks to compute and instead of computing them locally she sends these tasks to available "workers" that compute and report back the task results. We categorize these workers in three generic types: altruistic, malicious and rational. Altruistic workers that always return the correct result, malicious workers that always return an incorrect result, and rational workers that decide to reply or not truthfully depending on what increases their benefit. We design a reinforcement learning mechanism to induce a correct behavior to rational workers, while the mechanism is complemented by four reputation schemes that cope with malice. The goal of the mechanism is to reach a state of eventual correctness, that is, a stable state of the system in which the master always obtains the correct task results. Analysis of the system gives provable guarantees under which truthful behavior can be ensured. Finally, we observe the behavior of the mechanism through simulations that use realistic system parameters values. Simulations not only agree with the analysis but also reveal interesting trade-offs between various metrics and parameters. Finally, the four reputation schemes are assessed against the tolerance to cheaters.

cs.DC↗

arXiv2016

Superintelligence cannot be contained: Lessons from Computability Theory

Superintelligence is a hypothetical agent that possesses intelligence far surpassing that of the brightest and most gifted human minds. In light of recent advances in machine intelligence, a number of scientists, philosophers and technologists have revived the discussion about the potential catastrophic risks entailed by such an entity. In this article, we trace the origins and development of the neo-fear of superintelligence, and some of the major proposals for its containment. We argue that such containment is, in principle, impossible, due to fundamental limits inherent to computing itself. Assuming that a superintelligence will contain a program that includes all the programs that can be executed by a universal Turing machine on input potentially as complex as the state of the world, strict containment requires simulations of such a program, something theoretically (and practically) infeasible.

cs.CY↗

arXiv2014

Power-efficient Assignment of Virtual Machines to Physical Machines

Motivated by current trends in cloud computing, we study a version of the generalized assignment problem where a set of virtual processors has to be implemented by a set of identical processors. For literature consistency, we say that a set of virtual machines (VMs) is assigned to a set of physical machines (PMs). The optimization criteria is to minimize the power consumed by all the PMs. We term the problem Virtual Machine Assignment (VMA). Crucial differences with previous work include a variable number of PMs, that each VM must be assigned to exactly one PM (i.e., VMs cannot be implemented fractionally), and a minimum power consumption for each active PM. Such infrastructure may be strictly constrained in the number of PMs or in the PMs' capacity, depending on how costly (in terms of power consumption) is to add a new PM to the system or to heavily load some of the existing PMs. Low usage or ample budget yields models where PM capacity and/or the number of PMs may be assumed unbounded for all practical purposes. We study 4 VMA problems depending on whether the capacity or the number of PMs is bounded or not. Specifically, we study hardness and online competitiveness for a variety of cases. To the best of our knowledge, this is the first comprehensive study of the VMA problem for this cost function.

cs.DS↗

arXiv2014

A Measurement-based Analysis of the Energy Consumption of Data Center Servers

Energy consumption is a growing issue in data centers, impacting their economic viability and their public image. In this work we empirically characterize the power and energy consumed by different types of servers. In particular, in order to understand the behavior of their energy and power consumption, we perform measurements in different servers. In each of them, we exhaustively measure the power consumed by the CPU, the disk, and the network interface under different configurations, identifying the optimal operational levels. One interesting conclusion of our study is that the curve that defines the minimal CPU power as a function of the load is neither linear nor purely convex as has been previously assumed. Moreover, we find that the efficiency of the various server components can be maximized by tuning the CPU frequency and the number of active cores as a function of the system and network load, while the block size of I/O operations should be always maximized by applications. We also show how to estimate the energy consumed by an application as a function of some simple parameters, like the CPU load, and the disk and network activity. We validate the proposed approach by accurately estimating the energy of a map-reduce computation in a Hadoop platform.

cs.DC↗

arXiv2013

Algorithmic Mechanisms for Reliable Internet-based Computing under Collusion

In this work, using a game-theoretic approach, cost-sensitive mechanisms that lead to reliable Internet-based computing are designed. In particular, we consider Internet-based master-worker computations, where a master processor assigns, across the Internet, a computational task to a set of potentially untrusted worker processors and collects their responses. Workers may collude in order to increase their benefit. Several game-theoretic models that capture the nature of the problem are analyzed, and algorithmic mechanisms that, for each given set of cost and system parameters, achieve high reliability are designed. Additionally, two specific realistic system scenarios are studied. These scenarios are a system of volunteer computing like SETI, and a company that buys computing cycles from Internet computers and sells them to its customers in the form of a task- computation service. Notably, under certain conditions, non redundant allocation yields the best trade-off between cost and reliability.

cs.DC↗

arXiv2011

Probabilistic Bounds on the Length of a Longest Edge in Delaunay Graphs of Random Points in d-Dimensions

Motivated by low energy consumption in geographic routing in wireless networks, there has been recent interest in determining bounds on the length of edges in the Delaunay graph of randomly distributed points. Asymptotic results are known for random networks in planar domains. In this paper, we obtain upper and lower bounds that hold with parametric probability in any dimension, for points distributed uniformly at random in domains with and without boundary. The results obtained are asymptotically tight for all relevant values of such probability and constant number of dimensions, and show that the overhead produced by boundary nodes in the plane holds also for higher dimensions. To our knowledge, this is the first comprehensive study on the lengths of long edges in Delaunay graphs

cs.CG↗

arXiv2011

Fast Isomorphism Testing of Graphs with Regularly-Connected Components

The Graph Isomorphism problem has both theoretical and practical interest. In this paper we present an algorithm, called conauto-1.2, that efficiently tests whether two graphs are isomorphic, and finds an isomorphism if they are. This algorithm is an improved version of the algorithm conauto, which has been shown to be very fast for random graphs and several families of hard graphs. In this paper we establish a new theorem that allows, at very low cost, the easy discovery of many automorphisms. This result is especially suited for graphs with regularly connected components, and can be applied in any isomorphism testing and canonical labeling algorithm to drastically improve its performance. In particular, algorithm conauto-1.2 is obtained by the application of this result to conauto. The resulting algorithm preserves all the nice features of conauto, but drastically improves the testing of graphs with regularly connected components. We run extensive experiments, which show that the most popular algorithms (namely, nauty and bliss) can not compete with conauto-1.2 for these graph families.

cs.DS↗

arXiv2011

Opportunistic Information Dissemination in Mobile Ad-hoc Networks: adaptiveness vs. obliviousness and randomization vs. determinism

In this paper the problem of information dissemination in Mobile Ad-hoc Networks (MANET) is studied. The problem is to disseminate a piece of information, initially held by a distinguished source node, to all nodes in a set defined by some predicate. We use a model of MANETs that is well suited for dynamic networks and opportunistic communication. In this model nodes are placed in a plane, in which they can move with bounded speed, and communication between nodes occurs over a collision-prone single channel. In this setup informed and uninformed nodes can be disconnected for some time (bounded by a parameter alpha), but eventually some uninformed node must become neighbor of an informed node and remain so for some time (bounded by a parameter beta). In addition, nodes can start at different times, and they can crash and recover. Under the above framework, we show negative and positive results for different types of randomized protocols, and we put those results in perspective with respect to previous deterministic results.

cs.DS↗

arXiv2009

Performance of Random Walks in One-Hop Replication Networks

Random walks are gaining much attention from the networks research community. They are the basis of many proposals aimed to solve a variety of network-related problems such as resource location, network construction, nodes sampling, etc. This interest on random walks is justified by their inherent properties. They are very simple to implement as nodes only require local information to take routing decisions. Also, random walks demand little processing power and bandwidth. Besides, they are very resilient to changes on the network topology. Here, we quantify the effectiveness of random walks as a search mechanism in one-hop replication networks: networks where each node knows its neighbors' identity/resources, and so it can reply to queries on their behalf. Our model focuses on estimating the expected average search time of the random walk by applying network queuing theory. To do this, we must provide first the expected average search length. This is computed by means of estimations of the expected average coverage at each step of the random walk. This model takes into account the revisiting effect: the fact that, as the random walk progresses, the probability of arriving to nodes already visited increases, which impacts on how the network coverage evolves. That is, we do not model the coverage as a memoryless process. Furthermore, we conduct a series of simulations to evaluate, in practice, the above mentioned metrics. Our results show a very close correlation between the analytical and the experimental results.

cond-mat.dis-nn↗