We describe the multi-GPU gradient boosting algorithm implemented in the XGBoost library (https://github.com/dmlc/xgboost). Our algorithm allows fast, scalable training on multi-GPU systems with all of the features of the XGBoost library. We employ data compression techniques to minimise the usage of scarce GPU memory …
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
169,051 papers · 148 categories
Trend · papers per month
16 results for “multi-GPU”
Efficiently calibrates SABR/LIBOR models to real market caplets and swaptions data.
problem Calibration of stochastic volatility models to real market data.
method Proposes a parallelized simulated annealing algorithm for multi-GPUs.
result Numerical results show advantages of using multi-GPUs for SABR/LIBOR model calibration.
Exact GPs trained on over a million points in under 2 hours.
problem Computational limitations of exact Gaussian processes for large datasets.
method Multi-GPU parallelization and linear conjugate gradients for kernel matrix multiplication.
result Exact GPs can be trained on over a million points in less than 2 hours.
MLPerf benchmark suite evaluates diverse ML applications, highlighting system bottlenecks.
problem Understanding and optimizing ML applications across different models and systems.
method Analysis of MLPerf benchmark suite characteristics and comparison with previous benchmarks.
result Optimal distributed deep learning training requires dedicated low latency interconnects.
We propose a new algorithm called Parle for parallel training of deep networks that converges 2-4x faster than a data-parallel implementation of SGD, while achieving significantly improved error rates that are nearly state-of-the-art on several benchmarks including CIFAR-10 and CIFAR-100, without introducing any additi…
We describe a simple, low-level approach for embedding probabilistic programming in a deep learning ecosystem. In particular, we distill probabilistic programming down to a single abstraction---the random variable. Our lightweight implementation in TensorFlow enables numerous applications: a model-parallel variational …
Massively parallel architectures such as the GPU are becoming increasingly important due to the recent proliferation of data. In this paper, we propose a key class of hybrid parallel graphlet algorithms that leverages multiple CPUs and GPUs simultaneously for computing k-vertex induced subgraph statistics (called graph…
Novel IMEX scheme solves financial PDEs with mixed derivatives.
problem Numerical approximations for financial PDEs with mixed derivatives.
method Second order finite volume IMEX Runge-Kutta scheme.
result Achieves true second order convergence with non-regular initial conditions.
Training modern deep learning models requires large amounts of computation, often provided by GPUs. Scaling computation from one GPU to many can enable much faster training and research progress but entails two complications. First, the training library must support inter-GPU communication. Depending on the particular …
Distributed Reinforcement Learning of Targeted Grasping with Active Vision for Mobile Manipulatorscs.RO
Robotic grasping system learns to target objects from a single image.
problem Robotic grasping in unstructured environments.
method Distributed reinforcement learning, active vision, synchronous SGD.
result System learns to grasp unseen objects without retraining.
GeneCAI optimizes DNN compression hyper-parameters for mobile devices.
problem Efficient deployment of complex DNNs on resource-limited devices.
method GeneCAI uses genetic algorithm to learn optimal hyper-parameters.
result GeneCAI finds models with better accuracy-complexity trade-off.
This work optimizes deep learning training by combining data and model parallelism.
problem Training large models with multiple GPUs suffers from high communication overhead and statistical efficiency loss.
method Hybrid parallelization combining data and model parallelism.
result Hybrid training provides significant speedup compared to data parallelism alone.
The recent popularity of deep neural networks (DNNs) has generated a lot of research interest in performing DNN-related computation efficiently. However, the primary focus is usually very narrow and limited to (i) inference -- i.e. how to efficiently execute already trained models and (ii) image classification networks…
Deterministic Inference across Tensor Parallel Sizes That Eliminates Training-Inference Mismatchcs.LG
New method ensures consistent inference across different tensor parallel sizes for large language models.
problem Non-deterministic inference in large language models due to inconsistent reduction orders across GPUs.
method Tree-Based Invariant Kernels (TBIK) that align intra- and inter-GPU reduction orders through a unified hierarchical binary tree structure.
result Bit-wise identical results across different tensor parallel sizes for RL training.
Spectral learning extends matrix methods to tensors for better latent variable modeling.
problem Limitations of matrix-based spectral methods in capturing non-Gaussian data.
method Extend spectral decomposition to tensor-based methods for higher-order moments.
result Tensor decomposition can identify latent effects missed by matrix methods.
A new method for flow matching reduces computational costs and improves performance.
problem Efficiently matching flow models to target data distributions.
method Semidiscrete formulation of optimal transport (SD-OT) using SGD and maximum inner product search (MIPS).
result Semidiscrete FM (SD-FM) outperforms batch-OT and traditional flow matching methods.