arXiv subjects

Soojeong Kim

Publications and source records attributed to Soojeong Kim.

4 recordsLinked to original sources

arXiv2024

Minor Issues Escalated to Critical Levels in Large Samples: A Permutation-Based Fix

In the big data era, the need to reevaluate traditional statistical methods is paramount due to the challenges posed by vast datasets. While larger samples theoretically enhance accuracy and hypothesis testing power without increasing false positives, practical concerns about inflated Type-I errors persist. The prevalent belief is that larger samples can uncover subtle effects, necessitating dual consideration of p-value and effect size. Yet, the reliability of p-values from large samples remains debated. This paper warns that larger samples can exacerbate minor issues into significant errors, leading to false conclusions. Through our simulation study, we demonstrate how growing sample sizes amplify issues arising from two commonly encountered violations of model assumptions in real-world data and lead to incorrect decisions. This underscores the need for vigilant analytical approaches in the era of big data. In response, we introduce a permutation-based test to counterbalance the effects of sample size and assumption discrepancies by neutralizing them between actual and permuted data. We demonstrate that this approach effectively stabilizes nominal Type I error rates across various sample sizes, thereby ensuring robust statistical inferences even amidst breached conventional assumptions in big data. For reproducibility, our R codes are publicly available at: \url{https://github.com/ubcxzhang/bigDataIssue}.

stat.ME↗

arXiv2020

Accelerating Multi-Model Inference by Merging DNNs of Different Weights

Standardized DNN models that have been proved to perform well on machine learning tasks are widely used and often adopted as-is to solve downstream tasks, forming the transfer learning paradigm. However, when serving multiple instances of such DNN models from a cluster of GPU servers, existing techniques to improve GPU utilization such as batching are inapplicable because models often do not share weights due to fine-tuning. We propose NetFuse, a technique of merging multiple DNN models that share the same architecture but have different weights and different inputs. NetFuse is made possible by replacing operations with more general counterparts that allow a set of weights to be associated with only a certain set of inputs. Experiments on ResNet-50, ResNeXt-50, BERT, and XLNet show that NetFuse can speed up DNN inference time up to 3.6x on a NVIDIA V100 GPU, and up to 3.0x on a TITAN Xp GPU when merging 32 model instances, while only using up a small additional amount of GPU memory.

cs.LG↗

arXiv2019

Parallax: Sparsity-aware Data Parallel Training of Deep Neural Networks

The employment of high-performance servers and GPU accelerators for training deep neural network models have greatly accelerated recent advances in deep learning (DL). DL frameworks, such as TensorFlow, MXNet, and Caffe2, have emerged to assist DL researchers to train their models in a distributed manner. Although current DL frameworks scale well for image classification models, there remain opportunities for scalable distributed training on natural language processing (NLP) models. We found that current frameworks show relatively low scalability on training NLP models due to the lack of consideration to the difference in sparsity of model parameters. In this paper, we propose Parallax, a framework that optimizes data parallel training by utilizing the sparsity of model parameters. Parallax introduces a hybrid approach that combines Parameter Server and AllReduce architectures to optimize the amount of data transfer according to the sparsity. Experiments show that Parallax built atop TensorFlow achieves scalable training throughput on both dense and sparse models while requiring little effort from its users. Parallax achieves up to 2.8x, 6.02x speedup for NLP models than TensorFlow and Horovod with 48 GPUs, respectively. The training speed for the image classification models is equal to Horovod and 1.53x faster than TensorFlow.

cs.DC↗

arXiv2018

Improving the Expressiveness of Deep Learning Frameworks with Recursion

Recursive neural networks have widely been used by researchers to handle applications with recursively or hierarchically structured data. However, embedded control flow deep learning frameworks such as TensorFlow, Theano, Caffe2, and MXNet fail to efficiently represent and execute such neural networks, due to lack of support for recursion. In this paper, we add recursion to the programming model of existing frameworks by complementing their design with recursive execution of dataflow graphs as well as additional APIs for recursive definitions. Unlike iterative implementations, which can only understand the topological index of each node in recursive data structures, our recursive implementation is able to exploit the recursive relationships between nodes for efficient execution based on parallel computation. We present an implementation on TensorFlow and evaluation results with various recursive neural network models, showing that our recursive implementation not only conveys the recursive nature of recursive neural networks better than other implementations, but also uses given resources more effectively to reduce training and inference time.

cs.LG↗