FedADC Accelerated Federated Learning with Drift Control

Federated learning (FL) has become de facto framework for collaborativelearning among edge devices . The core of the FL strategy is the use of stochastic gradient descent (SGD) in a distributed manner . Largescale implementation of FL brings new challenges, such as the incorporation ofacceleration techniques designed for SGD into the distributed setting, and mitigation of the drift problem due to non-homogeneous distribution of localdatasets .…

Causality is Graphically Simple

Events in distributed systems include sending or receiving messages, or changing some state in a node . Not all events are related, but some events can influence how other, later events, occur . For instance, a reply to areceived mail message is influenced by that message, and maybe by other priormessages also received .…

Edge of the Earth Replication Optimizing Content Delivery in Large LEO Satellite Communication Networks

Large low earth orbit (LEO) satellite networks such as SpaceX’s Starlink promise to deliver low-latency, high-bandwidth Internet access with global coverage . They could potentially serve billions of Internet-connecteddevices . Replicating web content within satellites can reduce bandwidth use in the constellation by 93% over an approach without replicationin the network, while storing just 0.01% of all content in individualsatellites .…

Container Orchestration on HPC Systems

Torque-Operator (a plugin) serves as a bridge between HPC workload managers and container Orchestrators . Containers can encapsulate complex programs with theirdependencies in isolated environments . HPC clusters lack micro-services support and deeply integrated management, as opposed to container orchestrators (e.g.…

Online Service Migration in Edge Computing with Incomplete Information A Deep Recurrent Actor Critic Method

Multi-access Edge Computing (MEC) is a key technology in the fifth-generation(5G) network and beyond . Service migration needs to decide where to migrate user services for maintaining high Quality-of-Service(QoS) When users roam between MEC servers with limited coverage and capacity, finding an optimal migration policy is intractable due to the highly dynamic MEC environment and user mobility .…

Prizes Signal Scientific Revolutions

Scientific revolutions affect funding, investments, and technological advances . We investigated a possible signal predicting a topic’s revolutionarygrowth – its association with a scientific prize . We found that in the year following a prize, a topic experiences an onset of extraordinary growth in impact and talent that continues into the future .…

Listing Small Minimal Separators of a Graph

A minimal $a,b$-separator of $G$ is an inclusion-wise minimal vertices that separate $a$ and $b$ . We give an algorithm which enumerates such minimal separators, outputting the first $R$ minimalseparators in at most $poly(n) R \cdot \min(4^k, R)$ time . We also discuss barriers for obtaining a polynomial-delay algorithm .…

Indirect Identification of Horizontal Gene Transfer

Several implicit methods to infer Horizontal Gene Transfer (HGT) focus on pairs of genes that have diverged only after the divergence of the two species in which the genes reside . This situation defines the edge set of a graph, thelater-divergence-time (LDT) graph, whose vertices correspond to genes colored by their species .…

Clustering with Iterated Linear Optimization

We introduce a novel method for clustering using a semidefinite programming(SDP) relaxation of the Max k-Cut problem . The approach is based on a newmethodology for rounding the solution of an SDP using iterated linearoptimization . We show that usingfixed point iteration for rounding for rounding leads to significantly better results when compared to randomized rounding .…

Nearly tight Trotterization of interacting electrons

We consider simulating quantum systems on digital quantum computers . We use Trotterization for a class of interacting electrons that encompasses various physical systems . We estimate the simulation error by taking the transition amplitude ofnested commutators of Hamiltonian terms within the $eta$-electron manifold .…

Maximum 0 1 Timed Matching on Temporal Graphs

Temporal graphs are graphs where the topology and/or other properties of the graph change with time . They have been used to model applications with temporalinformation in various domains . Problems on static graphs become morechallenging to solve in temporal graphs because of dynamically changing topology .…

Programmable Quantum Annealers as Noisy Gibbs Samplers

Quantum annealing embodies a promising computationaligm that is intimately related to the complexity of energy landscapes inGibbs distributions . Drawing independent samples from high-dimensional probability distributions represents the major computational bottleneck for modern algorithms, including machine learning frameworks such as deep learning .…

Physical deep learning based on optimal control of dynamical systems

A central topic in recent artificial intelligence technologies is deeplearning, which can be regarded as a multilayer feedforward neural network . Here, we present a pattern recognitionbased on optimal control of continuous-time dynamical systems . The proposed approach enables to gain insight into mechanisms of deep networkprocessing in the framework of an optimal control problem and opens a novelpathway to realize physical computing hardware .…

DECOR GAN 3D Shape Detailization by Conditional Refinement

We introduce a deep generative network for 3D shape detailization, akin to stylization with the style being geometric details . The output shape preserves the overall structure (or content) of the input, while its detail generation is conditioned on an input “style code” We demonstrate that ourmethod can refine a coarse shape into a variety of detailed shapes .…

Temporal Graph Modeling for Skeleton based Action Recognition

Graph Convolutional Networks (GCNs) have obtained remarkable performance for skeleton-based action recognition . The proposed TE-GCN constructs temporal relationgraph to capture complex temporal dynamic . It is state-of-the-art performance by making contribution to temporal modeling for action recognition, says the proposed model achieves the state of theart performance .…

Object Centric Neural Scene Rendering

We present a method for composing photorealistic scenes from captured images of objects . Instead of learning a scene radiance field as a NeRF does, we propose to learn object-centric neural scattering functions (OSFs), arepresentation that models per-object light transport implicitly using alighting- and view-dependent neural network .…

Continuous Positional Payoffs

What payoffs are positional for (deterministic) two-player antagonistic games? In this paper we study this question for payoffsthat are continuous . We establish positionality of contracting payoffs via a Shapley’s fixed pointargument . Our main result states that any continuous positional payoff is acomposition of a non-decreasing continuous function and a contracting payoff .…

How the emotion s type and intensity affect rumor spreading

This paper sheds light on how DMs’emotional type and intensityaffect rumor spreading . Pessimism has a more significant influence on the stability of the evolutionary game thanoptimism . The government’s emotion types are more sensitive to the game resultsthan netizens, and the emotional intensity is proportional to the evolutionspeed of the game .…

Incentive Mechanism Design for Distributed Coded Machine Learning

A distributed machine learning platform needs to recruit many heterogeneous nodes to finish computation simultaneously . As a result, the overall performance may be degraded due to straggling workers . By introducingundundancy into computation, coded machine learning can effectively improve the runtime performance by recovering the final computation result through the first $k$ (out of the total $n$) workers who finish computation .…

Indecision Modeling

AI systems are often used to make or contribute to important decisions in a range of applications, including criminal justice, hiring, and medicine . Since these decisions impact human lives, it is important that the AIsystems act in ways which align with human values .…

Session based k NNs with Semantic Suggestions for Next item Prediction

Authors propose aconceptual (cSkNN) model extension for the next-item prediction . They use anNLP technique to parse salient concepts from the product titles to create product generalizations that are used for change detection and a recommendation list post-filtering . Authors say the extension is an improvement over the existing SkNN method on a sparse fashione-commerce dataset.…

Query expansion with artificially generated texts

A well-known way to improve the performance of document retrieval is to expand the user’s query . In this paper, we explore the use of text generation toautomatically expand the queries . This conceptually simple approach can easily be implemented on any IR system thanks to the availability of GPT code and models .…

Cache aided General Linear Function Retrieval

Coded Caching, proposed by Maddah-Ali and Niesen (MAN), has the potential to reduce network traffic by pre-storing content in the users’ local memories . This paper considers the linear function retrieval version of the original coded cachingsetting, where users are interested in retrieving a number of linearcombinations of the data points stored at the server, as opposed to a single file .…

Extracting Smart Contracts Tested and Verified in Coq

We implement extraction of Coq programs to functional languages based on MetaCoq’s certified erasure . We prove the pass correct wrt. a conventionalcall-by-value operational semantics of functional languages . We test several complex contracts such as a DAO-like contract, anescrow contract, and an implementation of a Decentralized Finance (DeFi) contract .…

A Novice Friendly Induction Tactic for Lean

In theorem provers based on dependent type theory such as Coq and Lean, induction tactics are omnipresent in proof scripts . Existing induction tactics do not reliably support inductive predicates and relations . This paper describes a new induction tactic, implemented in Lean 3, which addresses these issues .…

Learning Based Quality Assessment for Image Super Resolution

Image Super-Resolution (SR) techniques improve visual quality by enhancing spatial resolution of images . The resulting SR Image quality databasewith Semi-Automatic Ratings (SISAR) database, so far the largest of SR-IQA database,contains 8,400 images of 100 natural scenes . We train an end-to-end Deep ImageSR Quality (DISQ) model by employing two-stream Deep Neural Networks (DNNs) and a feature fusion network for qualityprediction .…

Optimizing the Parameters of A Physical Exercise Dose Response Model An Algorithmic Comparison

The purpose of this research was to compare the robustness and performance of a local and global optimization algorithm when given the task of fitting the parameters of a common non-linear dose-response model . The results of ourcomparison over 1000 experimental runs demonstrate the superior performance of the evolutionary computation based algorithm to consistently achieve a stronger model fit and holdout performance in comparison to the local search algorithm .…

Neural Pruning via Growing Regularization

Regularization has long been utilized to learn sparsity in deep neuralnetwork pruning . In this work, we extend its application to a new scenariowhere the regularization grows large gradually to tackle two central problemsof pruning: pruning schedule and weight importance scoring .…

Tag based Genetic Regulation for Genetic Programming

Tag-based genetic regulation extends existing tag-based naming schemes to allow programsto “promote” and “repress” code modules . This extension allows evolution to structure a program as a gene regulatory network where program modules areregulated based on instruction executions . We find thattag-based regulation improves problem-solving performance on context-dependent problems .…