RecD: Deduplication for End-to-End Deep Learning Recommendation Model Training Infrastructure
Authors:
Mark Zhao,
Dhruv Choudhary,
Devashish Tyagi,
Ajay Somani,
Max Kaplan,
Sung-Han Lin,
Sarunya Pumma,
Jongsoo Park,
Aarti Basant,
Niket Agarwal,
Carole-Jean Wu,
Christos Kozyrakis
Abstract:
We present RecD (Recommendation Deduplication), a suite of end-to-end infrastructure optimizations across the Deep Learning Recommendation Model (DLRM) training pipeline. RecD addresses immense storage, preprocessing, and training overheads caused by feature duplication inherent in industry-scale DLRM training datasets. Feature duplication arises because DLRM datasets are generated from interactio…
▽ More
We present RecD (Recommendation Deduplication), a suite of end-to-end infrastructure optimizations across the Deep Learning Recommendation Model (DLRM) training pipeline. RecD addresses immense storage, preprocessing, and training overheads caused by feature duplication inherent in industry-scale DLRM training datasets. Feature duplication arises because DLRM datasets are generated from interactions. While each user session can generate multiple training samples, many features' values do not change across these samples. We demonstrate how RecD exploits this property, end-to-end, across a deployed training pipeline. RecD optimizes data generation pipelines to decrease dataset storage and preprocessing resource demands and to maximize duplication within a training batch. RecD introduces a new tensor format, InverseKeyedJaggedTensors (IKJTs), to deduplicate feature values in each batch. We show how DLRM model architectures can leverage IKJTs to drastically increase training throughput. RecD improves the training and preprocessing throughput and storage efficiency by up to 2.48x, 1.79x, and 3.71x, respectively, in an industry-scale DLRM training system.
△ Less
Submitted 1 May, 2023; v1 submitted 9 November, 2022;
originally announced November 2022.
Implementation of Neural Network and feature extraction to classify ECG signals
Authors:
R Karthik,
Dhruv Tyagi,
Amogh Raut,
Soumya Saxena,
Rajesh Kumar M
Abstract:
This paper presents a suitable and efficient implementation of a feature extraction algorithm (Pan Tompkins algorithm) on electrocardiography (ECG) signals, for detection and classification of four cardiac diseases: Sleep Apnea, Arrhythmia, Supraventricular Arrhythmia and Long Term Atrial Fibrillation (AF) and differentiating them from the normal heart beat by using pan Tompkins RR detection follo…
▽ More
This paper presents a suitable and efficient implementation of a feature extraction algorithm (Pan Tompkins algorithm) on electrocardiography (ECG) signals, for detection and classification of four cardiac diseases: Sleep Apnea, Arrhythmia, Supraventricular Arrhythmia and Long Term Atrial Fibrillation (AF) and differentiating them from the normal heart beat by using pan Tompkins RR detection followed by feature extraction for classification purpose .The paper also presents a new approach towards signal classification using the existing neural networks classifiers.
△ Less
Submitted 17 February, 2018;
originally announced February 2018.
A note on dual demodulator continuous transmission frequency modulation technique
Authors:
Kapil Dev Tyagi,
Arun Kumar,
R. Bahl
Abstract:
The range resolution in conventional continuous time frequency modulation (CTFM) is inversely proportional to the signal bandwidth. The dual-demodulator continuous time frequency modulation (DD-CTFM) processing technique was proposed by Gough et al [1] as a method to increase the range resolution by making the output of DD-CTFM truly continuous. However, it has been found that in practice the rang…
▽ More
The range resolution in conventional continuous time frequency modulation (CTFM) is inversely proportional to the signal bandwidth. The dual-demodulator continuous time frequency modulation (DD-CTFM) processing technique was proposed by Gough et al [1] as a method to increase the range resolution by making the output of DD-CTFM truly continuous. However, it has been found that in practice the range resolution is still limited by the signal bandwidth. The limitation of DD-CTFM has been explained using simulations and mathematically in this paper.
△ Less
Submitted 11 January, 2017;
originally announced January 2017.