3d-SMRnet speeds up MPI system matrix recovery to 1 minute with high quality.
problem Slow system matrix recovery in MPI due to recalibration.
method 3d-System Matrix Recovery Network using deep learning.
result 3d-SMRnet recovers 3d system matrix with 64x subsampling in 1 minute.
Python framework for distributed Keras training on multiple GPUs/CPU.
problem Efficiently training neural networks on multiple GPUs/CPU.
method Built on Keras, uses MPI for coordination, suitable for supercomputing.
result Demonstrated performance on various system sizes.
MPI-FAUN tackles NMF for big data, offering scalable parallel algorithms.
problem Efficient parallel algorithms for NMF on big data.
method MPI-based framework for NMF, solving alternating NLS subproblems.
result Significant performance improvements over baseline implementations.
SparCML improves machine learning scalability by handling sparse data efficiently.
problem Scalability bottleneck in machine learning due to global allreduce operations.
method Design and implement communication protocols for sparse data, extending MPI.
result SparCML extends MPI to support sparse data, improving scalability.
AITuning uses reinforcement learning to optimize communication libraries.
problem Optimizing communication libraries for better performance in parallel applications.
method Deep reinforcement learning applied to tuning communication libraries.
result Demonstrates the effectiveness of reinforcement learning for tuning communication libraries.
Paper proposes a GPU-based system for training massive deep learning models in ads systems.
problem Training massive deep learning models with terabyte-scale parameters in ads systems.
method Hierarchical GPU parameter server with 3-layer storage (GPU High-Bandwidth Memory, CPU main memory, SSD).
result 4-node hierarchical GPU parameter server trains a model 2X faster than a 150-node in-memory system.
This paper considers a cross-layer adaptive modulation system that is modeled as a Markov decision process (MDP). We study how to utilize the monotonicity of the optimal transmission policy to relieve the computational complexity of dynamic programming (DP). In this system, a scheduler controls the bit rate of the m-qu…
Asynchronous SVI scales up Bayesian computation for massive data.
problem Scaling stochastic variational inference to massive data.
method Lock-free asynchronous parallel implementation of SVI.
result Linear speed-up and asymptotic convergence rate O ( 1 / ( T ) ) O(1/\sqrt(T)) O ( 1/ ( T )) . The paper optimizes dynamic scheduling for ring architectures in deep learning training.
problem Optimizing deep learning training times with ring architectures.
method Formulated a non-convex, non-linear, NP-hard integer programming problem and developed a doubling heuristic.
result Dynamic scheduling can significantly reduce job completion times in ring architectures.
Gradient coding improves distributed learning by tolerating failures.
problem Mitigating stragglers in distributed learning.
method Carefully replicating data blocks and coding across gradients.
result Shows tolerance to failures and stragglers for Synchronous Gradient Descent.
ParaMonte::Python streamlines Bayesian data analysis with fast Monte Carlo and MCMC routines.
problem Efficiently sampling posterior distributions in Bayesian modeling and data science.
method Serial and MPI-parallelized Markov Chain Monte Carlo (MCMC) routines.
result Automated model calibration and uncertainty quantification in Bayesian analysis.
Proposes a linear model for facial action recognition without requiring large datasets.
problem Limited annotated data for facial expression and action units.
method Exploits low-rank property across frames and group sparsity to subtract neutral faces and recognize actions.
result One-shot automatic method on raw face videos performs competitively and better than previous methods.
Financial institutions have massive computations to carry out overnight which are very demanding in terms of the consumed CPU. The challenge is to price many different products on a cluster-like architecture. We have used the Premia software to valuate the financial derivatives. In this work, we explain how Premia can …
New method speeds up nuclear-norm constrained learning over multiple machines.
problem Synchronization slowdown and high communication costs in large-scale learning.
method Asynchronous Stochastic Frank-Wolfe (SFW-asyn) method.
result SFW-asyn achieves the same convergence rate as vanilla SFW but with speed-ups almost linear to the number of machines.
ParaMonte simplifies Monte Carlo simulations for various scientific fields.
problem Efficiently performing Monte Carlo simulations for complex models.
method Unified, high-performance, parallelized library for C, C++, Fortran.
result Automates and streamlines Monte Carlo sampling for arbitrary-dimensional functions.
Combines soft greediness with Modified Policy Iteration for more efficient deep reinforcement learning.
problem Improving sample efficiency in deep reinforcement learning.
method Combines soft greediness with Modified Policy Iteration (MPI) for off-policy learning.
result The proposed algorithm is more sample efficient than original PPO and competitive with SAC.
Linear algebra approach for parallel deep learning models.
problem Training large DNNs in distributed environments.
method Linear algebraic approach to model parallelism.
result Manual development of backward operators for gradient-based training.
Etalumis bridges scientific simulators and probabilistic programming.
problem Infeasibility of rewriting scientific simulators for Bayesian inference.
method Cross-platform probabilistic execution protocol, MCMC and IC engines, distributed training of 3DCNN-LSTM.
result Achieved largest-scale posterior inference in a Turing-complete PPL for LHC use-case.
Enhanced SMC 2 ^2 2 uses gradients from CRN-PF in Langevin proposals for improved state and parameter estimation.
problem Challenges in high-dimensional parameter spaces for SMC 2 ^2 2 . method Leveraging gradients from a CRN-PF within a Langevin proposal.
result Higher effective sample size and more accurate parameter estimates.
Paper analyzes convergence of entropy-regularized reinforcement learning methods.
problem Ensuring sub-optimality control in entropy-regularized RL.
method Unified analysis of regularized MPI and VI schemes, providing convergence rates.
result Established sufficient conditions for convergence and provided explicit rates.
Study compares three performance metrics of Bangladeshi banks.
problem Comparing different performance metrics of commercial banks.
method Empirical framework using MPI, ROA, TSR over 2011-2015.
result Productivity growth was recorded, but profitability and stock performance were negative.
Enhanced 3D shape analysis using information geometry.
problem Challenges in comparing 3D point clouds due to their unstructured nature and complex geometry.
method Information geometric framework for 3D point cloud shape analysis using Gaussian Mixture Models (GMMs) on a statistical manifold. Proposed MSKL divergence with upper and lower bounds.
result MSKL provides stable and monotonically varying values that directly reflect geometric variation, outperforming traditional distances and existing KL approximations.
HybridSGD improves SGD performance by balancing computation and communication.
problem Limited scalability and performance of SGD due to communication costs.
method 2D parallel SGD method (HybridSGD) that trades off between 1D s s s -step SGD and 1D Federated SGD (FedAvg). result HybridSGD achieves better convergence than FedAvg at similar processor scales and up to 121x speedup over FedAvg.
JAMPI improves matrix multiplication in Spark, boosting performance by up to 24%.
problem Efficiently performing matrix multiplication in Spark.
method Combining asynchronous network IO, auto-vectorization, and barrier execution mode.
result Up to 24% performance increase in distributed matrix multiplication.
We present the first public release of our generic neural network training algorithm, called SkyNet. This efficient and robust machine learning tool is able to train large and deep feed-forward neural networks, including autoencoders, for use in a wide range of supervised and unsupervised learning applications, such as…
New DCD and BDCD methods for K-SVM and K-RR reduce communication costs.
problem Scalability issues in DCD and BDCD methods for kernel problems.
method Developed scalable s s s -step variants of DCD and BDCD methods. result Achieved strong scaling speedups of up to 9.8x over existing methods.
ParMAC optimizes nested functions in distributed systems, improving parallelism and speed.
problem Optimizing complex, nested machine learning models with large datasets and distributed computing.
method ParMAC introduces auxiliary coordinates for parallel training of submodels and coordinates, reducing communication overhead.
result ParMAC achieves high parallelism and low communication overhead, facilitating fast training of binary autoencoders.
Extension formulae on almost complex manifolds studied with applications.
problem Understanding almost complex manifolds through extension formulae.
method Provided extension formulae and decompositions for almost complex manifolds.
result Studied ( n , 0 ) (n,0) ( n , 0 ) -forms, ( n , 0 ) (n,0) ( n , 0 ) -Dolbeault cohomology group, and ( n , q ) (n,q) ( n , q ) -forms. The paper proves extension theorems for holomorphic sections from divisors.
problem Extension of holomorphic sections from reduced unions of strata of divisors.
method Proves an Ohsawa--Takegoshi type extension theorem.
result Qualitative results on extension from snc divisors and generic global generation of vector bundles.
Generalizes Nielsen equivalence theorem to hyperbolic group extensions.
problem Tackles Nielsen equivalence in hyperbolic group extensions.
method Generalizes a theorem by Juan Souto to a broader class of hyperbolic extensions.
result Includes all hyperbolic extensions of surfaces groups and free groups by Out$(F_n).
The paper connects group extensions, cochains, and spectral sequences.
problem Understanding the relationship between group extensions and spectral sequences.
method Using connection cochains, the paper derives a formula for the extension class.
result A formula clarifies the relation among connection cochains, extension classes, and the LHS spectral sequence.
Proves HNN extensions of nilpotent groups are left-orderable, constructs non-left-orderable examples.
problem Characterizing left-orderability in HNN extensions of groups.
method Analyzes HNN extensions of torsion-free nilpotent groups and left-orderable groups.
result Constructs examples of non-left-orderable HNN extensions of left-orderable groups.
Paper analyzes mathematical theory behind out-of-sample DR extensions.
problem Developing a solid mathematical foundation for out-of-sample DR extensions.
method Utilizes RKHS theory to treat DR extension as an extension of the identity on RKHS defined on X.
result Shows Nyström-type DR extension as an orthogonal projection and provides conditions for exact DR extension.
Examines differential smoothness in a specific skew PBW extension family.
problem Differential smoothness in skew PBW extensions.
method Investigates a specific family of skew PBW extensions.
result Results on differential smoothness of the family.
New insights into identifying mixtures of product distributions using Hadamard extensions.
problem Identifying mixtures of product distributions on binary variables.
method Analysis of Hadamard extensions of matrix products.
result Conditions for full column rank of Hadamard extensions.
A spacetime can be embedded in an enveloping space with all its extensions.
problem Existence and uniqueness of C0-maximal extensions in globally hyperbolic conformally flat spacetimes.
method Proving conformal embedding into an enveloping space containing all extensions.
result Existence and uniqueness of C0-maximal extensions proven.
We generalize the prequantization central extension of a group of diffeomorphisms preserving a closed 2-form ω(ω-invariant diffeomorphisms) to an abelian extension of a group of diffeomorphisms preserving a closed vector valued 2-form ω, up to a linear isomorphism (ω-equivariant diffeomorphisms). Every abelian extensio…
We give a new variant of L 2 L^2 L 2 -extension theorem for the jets of holomorphic sections and discuss the relation between the extension problem of singular Hermitian metrics with semipositive curvature.
Study on flux homomorphism and its extension in symplectic group of a disk.
problem Understanding the flux homomorphism and its extension in symplectic group.
method Defined and analyzed the flux homomorphism and its extension, determined the Euler class, and investigated its relation to group 2-cocycle and Calabi invariant.
result Determined the Euler class of the flux extension and investigated its relation to group 2-cocycle and Calabi invariant.
The purpose of this paper is to show how central extensions of (possibly infinite-dimensional) Lie algebras integrate to central extensions of étale Lie 2-groups. In finite dimensions, central extensions of Lie algebras integrate to central extensions of Lie groups, a fact which is due to the vanishing of π_2 for each …
We construct a Kruskal-Szekeres-type analytic extension of the Emparan-Reall black ring, and investigate its geometry. We prove that the extension is maximal, globally hyperbolic, and unique within a natural class of extensions. The key to those results is the proof that causal geodesics are either complete, or approac…
Analytic linearization and holomorphic extensions for proper groupoids.
problem Analytic linearization and holomorphic extensions of proper groupoids.
method Establish analytic linearization around invariant submanifolds and apply to holomorphic extensions.
result Proper groupoids admit holomorphic extensions.
We study the properties of Modified Riemann extensions evolving under Ricci flow. We obtain the necessary and sufficient condition for modified Riemann extension under Ricci flow to stay as modified Riemann extension. We also discuss the properties of the curvature tensors under Ricci flow.
The paper examines differential smoothness in skew PBW extensions over polynomial rings.
problem Differential smoothness in skew PBW extensions over polynomial rings.
method Investigation of skew PBW extensions over commutative polynomial rings.
result Results on differential smoothness for skew PBW extensions over polynomial rings.
Quasiregular maps get harmonic extensions in hyperbolic space.
problem Existence of harmonic extensions for quasiregular maps.
method Proving harmonic extensions for non-constant quasiregular maps on spheres.
result Non-constant quasiregular maps on spheres have harmonic extensions in hyperbolic space.
Simple construction of Lie 2-groups from loop group extensions.
problem Constructing Lie 2-groups from loop group extensions.
method Using conjugation action of loop group on its central extension.
result Simple construction of string 2-group as a strict Fréchet Lie 2-group.
Proves Girth Alternative for some HNN extensions, finds counterexamples.
problem Girth Alternative for HNN extensions of finitely generated groups.
method Proves Girth Alternative for a specific class of HNN extensions.
result Girth Alternative holds for some HNN extensions but fails in general.
Kan extensions help in data science extrapolation and learning.
problem Generalizing functions over larger sets in data science.
method Kan extensions in category theory applied to data science problems.
result Kan extensions can be used to derive classification and clustering algorithms.