arXiv research

Most distributed machine learning systems nowadays, including TensorFlow and CNTK, are built in a centralized fashion. One bottleneck of centralized algorithms lies on high communication cost on the central node. Motivated by this, we ask, can decentralized algorithms be faster than its centralized counterpart? Althoug…

Mime algorithm improves federated learning by adapting centralized methods.

The goal of this research project is to analyze the dynamics of social networks using machine learning techniques to locate maximal cliques and to find clusters for the purpose of identifying a target demographic. Unsupervised machine learning techniques are designed and implemented in this project to analyze a dataset…

Embedding graph nodes into a vector space can allow the use of machine learning to e.g. predict node classes, but the study of node embedding algorithms is immature compared to the natural language processing field because of a diverse nature of graphs. We examine the performance of node embedding algorithms with respe…

Federated learning has become increasingly important for modern machine learning, especially for data privacy-sensitive scenarios. Existing federated learning mostly adopts the central server-based architecture or centralized architecture. However, in many social network scenarios, centralized federated learning is not…

Algorithmic stablecoins optimize monetary policy to balance price stability.

Generalizes k-means to graphs using PageRank.

New algorithm detects cores in graphs with community structure, improving vertex selection for better clustering.

Stochastic gradient descent's long-term fluctuations are described by a diffusion limit.

I show that the solution of a standard clearing model commonly used in contagion analyses for financial systems can be expressed as a specific form of a generalized Katz centrality measure under conditions that correspond to a system-wide shock. This result provides a formal explanation for earlier empirical results wh…

Study analyzes a new algorithm for complex optimization problems.

This thesis tackles FL challenges with new methods and algorithms.

New algorithm reduces dimensionality in federated learning.

The exchange algorithm is studied for its convergence and asymptotic variance.

Communication is a key bottleneck in distributed training. Recently, an \emph{error-compensated} compression technology was particularly designed for the \emph{centralized} learning and receives huge successes, by showing significant advantages over state-of-the-art compression based methods in saving the communication…

Algorithm reduces regret in distributed kernel bandits with shared randomness.

We study the centralizer of a braid from the point of view of Garside theory, showing that generically a minimal set of generators can be computed very efficiently, as the ultra summit set of a generic braid has a very particular structure. We present an algorithm to compute the centralizer of a braid whose generic-cas…

Study improves accuracy of risk measures using advanced algorithms.

New method for zeroth-order stochastic gradient algorithms provides confidence intervals.

We give a new method to compute the centralizer of an element in Artin braid groups and, more generally, in Garside groups. This method, together with the solution of the conugacy problem given by the authors in a previous paper, are two main steps for solving conjugacy systems, thus breaking recently discovered crypto…

Vertex centrality measures are a multi-purpose analysis tool, commonly used in many application environments to retrieve information and unveil knowledge from the graphs and network structural properties. However, the algorithms of such metrics are expensive in terms of computational resources when running real-time ap…

We find a constructive bound for the word length of a generating set for the centralizer of an element of the Mapping Class Group. As a consequence, we show that it is algorithmically decidable whether two postcritically finite branched coverings of the sphere are Thurston equivalent.

The article considers the problem of existence and uniqueness of centrally symmetrical convex body for which the projection curvature radius function coincides with a given flag function. A necessary and sufficient condition is found that ensures a positive answer. An algorithm for construction the body in question is …

Study finds central points of double heptagon surface are not connection points.

Sensitive statistics are often collected across sets of users, with repeated collection of reports done over time. For example, trends in users' private preferences or software usage may be monitored via such reports. We study the collection of such statistics in the local differential privacy (LDP) model, and describe…

New SAGA algorithm with decreasing step for stochastic optimization.

The speed with which a learning algorithm converges as it is presented with more data is a central problem in machine learning --- a fast rate of convergence means less data is needed for the same level of performance. The pursuit of fast rates in online and statistical learning has led to the discovery of many conditi…

Study efficient algorithms for one-shot federated conformal prediction.

New centrality-based graph shift operators improve graph neural networks.

Decentralized learning achieves centralized performance via Gibbs measures.

Quantum computing aids in optimizing currency reserves for central banks.

We propose a communicationally and computationally efficient algorithm for high-dimensional distributed sparse learning. At each iteration, local machines compute the gradient on local data and the master machine solves one shifted $l_1$ regularized minimization problem. The communication cost is reduced from constant …

The paper analyzes Q-learning convergence rates with asynchronous updates.

New algorithm reduces regret in asynchronous multiplayer bandits to constant or logarithmic levels.

The paper establishes CLTs for Markov chains and improves sampling algorithms for heavy-tailed distributions.

A privacy-preserving algorithm for high-dimensional bandits.

Many functions of interest are in a high-dimensional space but exhibit low-dimensional structures. This paper studies regression of a $s$-Hölder function $f$ in $\mathbb{R}^D$ which varies along a central subspace of dimension $d$ while $d\ll D$. A direct approximation of $f$ in $\mathbb{R}^D$ with an $\varepsilon$ acc…

Training generative models like Generative Adversarial Network (GAN) is challenging for noisy data. A novel curriculum learning algorithm pertaining to clustering is proposed to address this issue in this paper. The curriculum construction is based on the centrality of underlying clusters in data points. The data point…

Distributed learning techniques such as federated learning have enabled multiple workers to train machine learning models together to reduce the overall training time. However, current distributed training algorithms (centralized or decentralized) suffer from the communication bottleneck on multiple low-bandwidth worke…

Study finds recurring patterns in cryptocurrency volatility and liquidity.

Introduces a reduction system for Artin-Tits groups, improving algorithms and proving periodicity results.

Solved a specific case of Salter's question on Burau representation.

In many signal processing and machine learning applications, datasets containing private information are held at different locations, requiring the development of distributed privacy-preserving algorithms. Tensor and matrix factorizations are key components of many processing pipelines. In the distributed setting, diff…

A novel decentralized deep learning algorithm using gradient-based optimization.

Paper improves CLT and bootstrap approximations for LSA with decreasing step size.

A new model detects complex network communities using node attributes.

Second-order guarantees for federated learning algorithms.

Stochastic gradient descent in continuous time (SGDCT) provides a computationally efficient method for the statistical learning of continuous-time models, which are widely used in science, engineering, and finance. The SGDCT algorithm follows a (noisy) descent direction along a continuous stream of data. The parameter …