اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدFundamental Limits of Nonintrusive Load Monitoring
62
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Provided an arbitrary nonintrusive load monitoring (NILM) algorithm, we seek bounds on the probability of distinguishing between scenarios, given an aggregate power consumption signal. We introduce a framework for studying a general NILM algorithm, and analyze the theory in the general case. Then, we specialize to the case where the error is Gaussian. In both cases, we are able to derive upper bounds on the probability of distinguishing scenarios. Finally, we apply the results to real data to derive bounds on the probability of distinguishing between scenarios as a function of the measurement noise, the sampling rate, and the device usage.
قيم البحث
اقرأ أيضاً
In general coding theory, we often assume that error is observed in transferring or storing encoded symbols, while the process of encoding itself is error-free. Motivated by recent applications of coding theory, in this paper, we consider the case wh
ere the process of encoding is distributed and prone to error. We introduce the problem of distributed encoding, comprising of $Kinmathbb{N}$ isolated source nodes and $Ninmathbb{N}$ encoding nodes. Each source node has one symbol from a finite field and sends it to all encoding nodes. Each encoding node stores an encoded symbol, as a function of the received symbols. However, some of the source nodes are controlled by the adversary and may send different symbols to different encoding nodes. Depending on the number of adversarial nodes, denoted by $betainmathbb{N}$, and the number of symbols that each one generates, denoted by $vinmathbb{N}$, the process of decoding from the encoded symbols could be impossible. Assume that a decoder connects to an arbitrary subset of $t inmathbb{N}$ encoding nodes and wants to decode the symbols of the honest nodes correctly, without necessarily identifying the sets of honest and adversarial nodes. In this paper, we study $t^*inmathbb{N}$, the minimum of $t$, which is a function of $K$, $N$, $beta$, and $v$. We show that when the encoding nodes use linear coding, $t^*_{textrm{linear}}=K+2beta(v-1)$, if $Nge K+2beta(v-1)$, and $t^*_{textrm{linear}}=N$, if $Nle K+2beta(v-1)$. In order to achieve $t^*_{textrm{linear}}$, we use random linear coding and show that in any feasible solution that the decoder finds, the messages of the honest nodes are decoded correctly. For the converse of the fundamental limit, we show that when the adversary behaves in a particular way, it can always confuse the decoder between two feasible solutions that differ in the message of at least one honest node.
In Quantum Illumination (QI), a signal beam initially entangled with an idler beam held at the receiver interrogates a target region bathed in thermal background light. The returned beam is measured jointly with the idler in order to determine whethe
r a weakly reflecting target is present. Using tools from quantum information theory, we derive lower bounds on the average error probability of detecting both specular and fading targets and on the mean squared error of estimating the reflectance of a detected target, which are obeyed by any QI transmitter satisfying a signal energy constraint. For bright thermal backgrounds, we show that the QI system using multiple copies of low-brightness two-mode squeezed vacuum states is nearly optimal. More generally, our results place limits on the best possible performance achievable using QI systems at all wavelengths, and at all signal and background noise levels.
It has been established that when the gradient coding problem is distributed among $n$ servers, the computation load (number of stored data partitions) of each worker is at least $s+1$ in order to resists $s$ stragglers. This scheme incurs a large ov
erhead when the number of stragglers $s$ is large. In this paper, we focus on a new framework called emph{approximate gradient coding} to mitigate stragglers in distributed learning. We show that, to exactly recover the gradient with high probability, the computation load is lower bounded by $O(log(n)/log(n/s))$. We also propose a code that exactly matches such lower bound. We identify a fundamental three-fold tradeoff for any approximate gradient coding scheme $dgeq O(log(1/epsilon)/log(n/s))$, where $d$ is the computation load, $epsilon$ is the error of gradient. We give an explicit code construction based on random edge removal process that achieves the derived tradeoff. We implement our schemes and demonstrate the advantage of the approaches over the current fastest gradient coding strategies.
Optimal caching of files in a content distribution network (CDN) is a problem of fundamental and growing commercial interest. Although many different caching algorithms are in use today, the fundamental performance limits of network caching algorithm
s from an online learning point-of-view remain poorly understood to date. In this paper, we resolve this question in the following two settings: (1) a single user connected to a single cache, and (2) a set of users and a set of caches interconnected through a bipartite network. Recently, an online gradient-based coded caching policy was shown to enjoy sub-linear regret. However, due to the lack of known regret lower bounds, the question of the optimality of the proposed policy was left open. In this paper, we settle this question by deriving tight non-asymptotic regret lower bounds in both of the above settings. In addition to that, we propose a new Follow-the-Perturbed-Leader-based uncoded caching policy with near-optimal regret. Technically, the lower-bounds are obtained by relating the online caching problem to the classic probabilistic paradigm of balls-into-bins. Our proofs make extensive use of a new result on the expected load in the most populated half of the bins, which might also be of independent interest. We evaluate the performance of the caching policies by experimenting with the popular MovieLens dataset and conclude the paper with design recommendations and a list of open problems.
Short term load forecasts will play a key role in the implementation of smart electricity grids. They are required to optimise a wide range of potential network solutions on the low voltage (LV) grid, including integrating low carbon technologies (su
ch as photovoltaics) and utilising battery storage devices. Despite the need for accurate LV level load forecasts, previous studies have mostly focused on forecasting at the individual household or building level using data from smart meters. In this study we provide detailed analysis of a variety of methods in terms of both point and probabilistic forecasting accuracy using data from 100 real LV feeders. Moreover, we investigate the effect of temperature (both actual and forecasts) on the accuracy of load forecasts. We present some important results on the drivers of LV forecasting accuracy that are crucial for the management of LV networks, along with an empirical comparison of forecast measures.
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
