亚洲男人的天堂2018av,欧美草比,久久久久久免费视频精选,国色天香在线看免费,久久久久亚洲av成人片仓井空

Adaptive gradient methods have shown excellent performances for solving many machine learning problems. Although multiple adaptive methods were recently studied, they mainly focus on either empirical or theoretical aspects and also only work for specific problems by using some specific adaptive learning rates. It is desired to design a universal framework for practical algorithms of adaptive gradients with theoretical guarantee to solve general problems. To fill this gap, we propose a faster and universal framework of adaptive gradients (i.e., SUPER-ADAM) by introducing a universal adaptive matrix that includes most existing adaptive gradient forms. Moreover, our framework can flexibly integrate the momentum and variance reduced techniques. In particular, our novel framework provides the convergence analysis support for adaptive gradient methods under the nonconvex setting. In theoretical analysis, we prove that our SUPER-ADAM algorithm can achieve the best known gradient (i.e., stochastic first-order oracle (SFO)) complexity of $\tilde{O}(\epsilon^{-3})$ for finding an $\epsilon$-stationary point of nonconvex optimization, which matches the lower bound for stochastic smooth nonconvex optimization. In numerical experiments, we employ various deep learning tasks to validate that our algorithm consistently outperforms the existing adaptive algorithms. Code is available at //github.com/LIJUNYI95/SuperAdam

Clustering is an important task with applications in many fields of computer science. We study the fully dynamic setting in which we want to maintain good clusters efficiently when input points (from a metric space) can be inserted and deleted. Many clustering problems are $\mathsf{APX}$-hard but admit polynomial time $O(1)$-approximation algorithms. Thus, it is a natural question whether we can maintain $O(1)$-approximate solutions for them in subpolynomial update time, against adaptive and oblivious adversaries. Only a few results are known that give partial answers to this question. There are dynamic algorithms for $k$-center, $k$-means, and $k$-median that maintain constant factor approximations in expected $\tilde{O}(k^{2})$ update time against an oblivious adversary. However, for these problems there are no algorithms known with an update time that is subpolynomial in $k$, and against an adaptive adversary there are even no (non-trivial) dynamic algorithms known at all. In this paper, we complete the picture of the question above for all these clustering problems. 1. We show that there is no fully dynamic $O(1)$-approximation algorithm for any of the classic clustering problems above with an update time in $n^{o(1)}h(k)$ against an adaptive adversary, for an arbitrary function $h$. 2. We give a lower bound of $\Omega(k)$ on the update time for each of the above problems, even against an oblivious adversary. 3. We give the first $O(1)$-approximate fully dynamic algorithms for $k$-sum-of-radii and for $k$-sum-of-diameters with expected update time of $\tilde{O}(k^{O(1)})$ against an oblivious adversary. 4. Finally, for $k$-center we present a fully dynamic $(6+\epsilon)$-approximation algorithm with an expected update time of $\tilde{O}(k)$ against an oblivious adversary.

We present the first formal verification of approximation algorithms for NP-complete optimization problems: vertex cover, independent set, set cover, center selection, load balancing, and bin packing. We uncover incompletenesses in existing proofs and improve the approximation ratio in one case. All proofs are uniformly invariant based.

We prove that every simple 2-connected subcubic graph on $n$ vertices with $n_2$ vertices of degree 2 has a TSP walk of length at most $\frac{5n+n_2}{4}-1$, confirming a conjecture of Dvo\v{r}\'ak, Kr\'al', and Mohar. This bound is best possible; there are infinitely many subcubic and cubic graphs whose minimum TSP walks have lengths $\frac{5n+n_2}{4}-1$ and $\frac{5n}{4} - 2$ respectively. We characterize the extremal subcubic examples meeting this bound. We also give a quadratic-time combinatorial algorithm for finding such a TSP walk. In particular, we obtain a $\frac{5}{4}$-approximation algorithm for the graphic TSP on simple cubic graphs, improving on the previously best known approximation ratio of $\frac{9}{7}$.

In this paper we present an invariance proof of three properties on Simpson's 4-slot algorithm, i.e. data-race freedom, data coherence and data freshness, which together implies linearisability of the algorithm. It is an extension of previous works whose proof focuses mostly on data-race freedom. In addition, our proof uses simply inductive invariants and transition invariants, whereas previous work uses more sophisticated machinery like separation logics, rely-guarantee or ownership transfer.

The Dyck language, which consists of well-balanced sequences of parentheses, is one of the most fundamental context-free languages. The Dyck edit distance quantifies the number of edits (character insertions, deletions, and substitutions) required to make a given parenthesis sequence well-balanced. RNA Folding involves a similar problem, where a closing parenthesis can match an opening parenthesis of the same type irrespective of their ordering. For example, in RNA Folding, both $\tt{()}$ and $\tt{)(}$ are valid matches, whereas the Dyck language only allows $\tt{()}$ as a match. Using fast matrix multiplication, it is possible to compute their exact solutions of both problems in time $O(n^{2.824})$. Whereas combinatorial algorithms would be more desirable, the two problems are known to be at least as hard as Boolean matrix multiplication. In terms of fast approximation algorithms that are combinatorial in nature, both problems admit an $\epsilon n$-additive approximation in $\tilde{O}(\frac{n^2}{\epsilon})$ time. Further, there is a $O(\log n)$-factor approximation algorithm for Dyck edit distance in near-linear time. In this paper, we design a constant-factor approximation algorithm for Dyck edit distance that runs in $O(n^{1.971})$ time. Moreover, we develop a $(1+\epsilon)$-factor approximation algorithm running in $\tilde{O}(\frac{n^2}{\epsilon})$ time, which improves upon the earlier additive approximation. Finally, we design a $(3+\epsilon)$-approximation that takes $\tilde{O}(\frac{nd}{\epsilon})$ time, where $d\ge 1$ is an upper bound on the sought distance. As for RNA folding, for any $s\ge1$, we design a factor-$s$ approximation algorithm that runs in $O(n+(\frac{n}{s})^3)$ time. To the best of our knowledge, this is the first nontrivial approximation algorithm for RNA Folding that can go below the $n^2$ barrier. All our algorithms are combinatorial.

We study fractional variants of the quasi-norms introduced by Brezis, Van Schaftingen, and Yung in the study of the Sobolev space $\dot W^{1,p}$. The resulting spaces are identified as a special class of real interpolation spaces of Sobolev-Slobodecki\u{\i} spaces. We establish the equivalence between Fourier analytic definitions and definitions via difference operators acting on measurable functions. We prove various new results on embeddings and non-embeddings, and give applications to harmonic and caloric extensions. For suitable wavelet bases we obtain a characterization of the approximation spaces for best $n$-term approximation from a wavelet basis via smoothness conditions on the function; this extends a classical result by DeVore, Jawerth and Popov.

Let $P$ be a set of points in $\mathbb{R}^d$, where each point $p\in P$ has an associated transmission range $\rho(p)$. The range assignment $\rho$ induces a directed communication graph $\mathcal{G}_{\rho}(P)$ on $P$, which contains an edge $(p,q)$ iff $|pq| \leq \rho(p)$. In the broadcast range-assignment problem, the goal is to assign the ranges such that $\mathcal{G}_{\rho}(P)$ contains an arborescence rooted at a designated node and whose cost $\sum_{p \in P} \rho(p)^2$ is minimized. We study trade-offs between the stability of the solution -- the number of ranges that are modified when a point is inserted into or deleted from $P$ -- and its approximation ratio. We introduce $k$-stable algorithms, which are algorithms that modify the range of at most $k$ points when they update the solution. We also introduce the concept of a stable approximation scheme (SAS). A SAS is an update algorithm that, for any given fixed parameter $\varepsilon>0$, is $k(\epsilon)$-stable and maintains a solution with approximation ratio $1+\varepsilon$, where the stability parameter $k(\varepsilon)$ only depends on $\varepsilon$ and not on the size of $P$. We study such trade-offs in three settings. - In $\mathbb{R}^1$, we present a SAS with $k(\varepsilon)=O(1/\varepsilon)$, which we show is tight in the worst case. We also present a 1-stable $(6+2\sqrt{5})$-approximation algorithm, a $2$-stable 2-approximation algorithm, and a $3$-stable $1.97$-approximation algorithm. - In $\mathbb{S}^1$ (where the underlying space is a circle) we prove that no SAS exists, even though an optimal solution can always be obtained by cutting the circle at an appropriate point and solving the resulting problem in $\mathbb{R}^1$. - In $\mathbb{R}^2$, we also prove that no SAS exists, and we present a $O(1)$-stable $O(1)$-approximation algorithm.

In recent years the sleeping model came to the focus of researchers. In this model nodes can go into a sleep state in which they spend no energy but at the same time cannot receive or send messages, nor can they perform internal computations. This model captures energy considerations of a problem. A problem P is an O-LOCAL problem if, given an acyclic orientation on the edges of the input graph, one can solve the problem as follows. Each vertex awaits the decisions of its parents according to the given orientation and can make its own decision in regard to P using only the information about its parents decisions. problems and showed that for this class of problems there is a deterministic algorithm that runs in $O(\log \Delta)$ awake time. The clock round complexity of that algorithm is $O(\Delta^2)$. In this work we offer three algorithms for the bf O-LOCAL class of problems with a trade off between awake complexity and clock round complexity. One of these algorithms requires only $O(\Delta^{1+\epsilon})$ clock rounds for some constant $\epsilon>0$ but still only $O(\log \Delta)$ awake time which improves on the algorithm in \cite{BM21}. We add to this two other algorithms that trade a higher awake complexity for lower clock round complexity. We note that the awake time incurred is not that significant. We offer dynamic algorithms in the sleeping model. We show three algorithms for solving dynamic problems in the O-LOCAL class as well as an algorithm for solving any dynamic decidable problem. We show that one can solve any {\bf O-LOCAL} problem in constant awake time in graphs with constant neighborhood independence. Specifically, our algorithm requires $O(K)$ awake time where $K$ is the neighborhood independence of the input graph. Graphs with bounded neighborhood independence are well studied with several results in recent years for several core problem in the distributed setting.

We consider the task of learning the parameters of a {\em single} component of a mixture model, for the case when we are given {\em side information} about that component, we call this the "search problem" in mixture models. We would like to solve this with computational and sample complexity lower than solving the overall original problem, where one learns parameters of all components. Our main contributions are the development of a simple but general model for the notion of side information, and a corresponding simple matrix-based algorithm for solving the search problem in this general setting. We then specialize this model and algorithm to four common scenarios: Gaussian mixture models, LDA topic models, subspace clustering, and mixed linear regression. For each one of these we show that if (and only if) the side information is informative, we obtain parameter estimates with greater accuracy, and also improved computation complexity than existing moment based mixture model algorithms (e.g. tensor methods). We also illustrate several natural ways one can obtain such side information, for specific problem instances. Our experiments on real data sets (NY Times, Yelp, BSDS500) further demonstrate the practicality of our algorithms showing significant improvement in runtime and accuracy.