scholarly journals Hindsight logging for model training

2020 ◽  
Vol 14 (4) ◽  
pp. 682-693
Author(s):  
Rolando Garcia ◽  
Eric Liu ◽  
Vikram Sreekanti ◽  
Bobby Yan ◽  
Anusha Dandamudi ◽  
...  
Keyword(s):  
Training Model ◽  
Database Systems ◽  
Slowing Down ◽  
Machine Learning Model ◽  
Experimental Process ◽  
Free Mode ◽  
Model Training ◽  
Post Hoc ◽  
Multiple Variants ◽  
Modern Machine

In modern Machine Learning, model training is an iterative, experimental process that can consume enormous computation resources and developer time. To aid in that process, experienced model developers log and visualize program variables during training runs. Exhaustive logging of all variables is infeasible, so developers are left to choose between slowing down training via extensive conservative logging, or letting training run fast via minimalist optimistic logging that may omit key information. As a compromise, optimistic logging can be accompanied by program checkpoints; this allows developers to add log statements post-hoc, and "replay" desired log statements from checkpoint---a process we refer to as hindsight logging. Unfortunately, hindsight logging raises tricky problems in data management and software engineering. Done poorly, hindsight logging can waste resources and generate technical debt embodied in multiple variants of training code. In this paper, we present methodologies for efficient and effective logging practices for model training, with a focus on techniques for hindsight logging. Our goal is for experienced model developers to learn and adopt these practices. To make this easier, we provide an open-source suite of tools for Fast Low-Overhead Recovery (flor) that embodies our design across three tasks: (i) efficient background logging in Python, (ii) adaptive periodic checkpointing, and (iii) an instrumentation library that codifies hindsight logging for efficient and automatic record-replay of model-training. Model developers can use each flor tool separately as they see fit, or they can use flor in hands-free mode, entrusting it to instrument their code end-to-end for efficient record-replay. Our solutions leverage techniques from physiological transaction logs and recovery in database systems. Evaluations on modern ML benchmarks demonstrate that flor can produce fast checkpointing with small user-specifiable overheads (e.g. 7%), and still provide hindsight log replay times orders of magnitude faster than restarting training from scratch.

scholarly journals UAV Recognition Based on Micro-Doppler Dynamic Attribute-Guided Augmentation Algorithm

2021 ◽  
Vol 13 (6) ◽  
pp. 1205
Author(s):  
Caidan Zhao ◽  
Gege Luo ◽  
Yilin Wang ◽  
Caiyun Chen ◽  
Zhiqiang Wu
Keyword(s):  
Recognition Performance ◽  
Principal Component ◽  
Texture Features ◽  
Training Model ◽  
Recognition Algorithm ◽  
Discrete Wavelet ◽  
Dynamic Feature ◽  
Complex Environments ◽  
Time Frequency ◽  
Model Training

A micro-Doppler signature (m-DS) based on the rotation of drone blades is an effective way to detect and identify small drones. Deep-learning-based recognition algorithms can achieve higher recognition performance, but they needs a large amount of sample data to train models. In addition to the hovering state, the signal samples of small unmanned aerial vehicles (UAVs) should also include flight dynamics, such as vertical, pitch, forward and backward, roll, lateral, and yaw. However, it is difficult to collect all dynamic UAV signal samples under actual flight conditions, and these dynamic flight characteristics will lead to the deviation of the original features, thus affecting the performance of the recognizer. In this paper, we propose a small UAV m-DS recognition algorithm based on dynamic feature enhancement. We extract the combined principal component analysis and discrete wavelet transform (PCA-DWT) time–frequency characteristics and texture features of the UAV’s micro-Doppler signal and use a dynamic attribute-guided augmentation (DAGA) algorithm to expand the feature domain for model training to achieve an adaptive, accurate, and efficient multiclass recognition model in complex environments. After the training model is stable, the average recognition accuracy rate can reach 98% during dynamic flight.

scholarly journals Enhancement of Multilayer Perceptron Model Training Accuracy through the Optimization of Hyperparameters: A Case Study of the Quality Prediction of Injection Molded Parts

2021 ◽  
Author(s):  
Kun-Cheng Ke ◽  
Ming-Shyan Huang
Keyword(s):  
Gradient Descent ◽  
Training Model ◽  
Learning Rate ◽  
Stochastic Gradient Descent ◽  
Activation Functions ◽  
Molded Parts ◽  
Injection Molded ◽  
Testing Accuracy ◽  
Model Training

Abstract Injection molding has been broadly used in the mass production of plastic parts and must meet the requirements of efficiency and quality consistency. Machine learning can effectively predict the quality of injection molded part. However, the performance of machine learning models largely depends on the accuracy of the training. Hyperparameters such as activation functions, momentum, and learning rate are crucial to the accuracy and efficiency of model training. This research further analyzed the influence of hyperparameters on testing accuracy, explored the corresponding optimal learning rate, and provided the optimal training model for predicting the quality of injection molded parts. In this study, stochastic gradient descent (SGD) and stochastic gradient descent with momentum were used to optimize the artificial neural network model. Through optimization of these training model hyperparameters, the width testing accuracy of the injection product improved. The experimental results indicated that in the absence of momentum effects, all five activation functions can achieve more than 90% of the training accuracy with a learning rate of 0.1. Moreover, when optimized with the SGD, the learning rate of the Sigmoid activation function was 0.1, and the testing accuracy reached 95.8%. Although momentum had the least influence on accuracy, it affected the convergence speed of the Sigmoid function, which reduced the number of required learning iterations (82.4% reduction rate). Optimizing hyperparameter settings can improve the accuracy of model testing and markedly reduce training time.

scholarly journals Recognition of Voltage Sag Sources Based on Phase Space Reconstruction and Improved VGG Transfer Learning

Entropy ◽  
2019 ◽  
Vol 21 (10) ◽  
pp. 999
Author(s):  
Yuting Pu ◽  
Honggeng Yang ◽  
Xiaoyang Ma ◽  
Xiangxun Sun
Keyword(s):  
Phase Space ◽  
Transfer Learning ◽  
Simulation Analysis ◽  
Recognition Accuracy ◽  
Training Model ◽  
Phase Space Reconstruction ◽  
Voltage Sag ◽  
Recognition Method ◽  
Model Training ◽  
Visual Geometry

The recognition of the voltage sag sources is the basis for formulating a voltage sag governance plan and clarifying the responsibility for the accident. Aiming at the recognition problem of voltage sag sources, a recognition method of voltage sag sources based on phase space reconstruction and improved Visual Geometry Group (VGG) transfer learning is proposed from the perspective of image classification. Firstly, phase space reconstruction technology is used to transform voltage sag signals, generate reconstruction images of voltage sag, and analyze the intuitive characteristics of different sag sources from reconstruction images. Secondly, combined with the attention mechanism, the standard VGG 16 model is improved to extract the features completely and prevent over-fitting. Finally, VGG transfer learning model uses the idea of transfer learning for training, which improves the efficiency of model training and the recognition accuracy of sag sources. The purpose of the training model is to minimize the cross entropy loss function. The simulation analysis verifies the effectiveness and superiority of the proposed method.

scholarly journals Calorie Detection of Food Image based on SVM Algorithm

2021 ◽  
Vol 9 (VI) ◽  
pp. 1836-1840
Author(s):  
Madhubala Kamble
Keyword(s):  
Machine Learning ◽  
Support Vector Machine ◽  
Training Model ◽  
Support Vector ◽  
Prototype System ◽  
Training Module ◽  
Text Data ◽  
Server Side ◽  
Machine Learning Model ◽  
Unique System

Nowadays, standard intake of healthy food is vital for keeping a diet to avoid obesity within the human body . In this paper, we present a totally unique system supported machine learning that automatically performs accurate classification of food images and estimates food attributes. This paper proposes a machine learning model consisting of a support vector machine that classifies food into specific categories within the training a part of the prototype system. The most purpose of the proposed method is to reinforce the accuracy of the pre-training model. The paper designs a prototype system supported the client server network model. The client sends an image detection request and processes it on the server side. The prototype system is meant with three main software components, including a pre-trained support vector machine training module for classification purposes, a text data training module for attribute estimation models, and a server-side module. We experimented with a selection of food categories, each containing thousands of images, and therefore the machine learning training to understand higher classification accuracy.

scholarly journals Training health cadres to response the COVID-19 pandemic in Pujon Kidul Village Malang Regency

2021 ◽  
Vol 2 (3) ◽  
pp. 140-145
Author(s):  
Yoyok Bekti Prasetyo ◽  
Nurul Zuriah ◽  
Joko Susilo
Keyword(s):  
Community Service ◽  
Task Force ◽  
Training Model ◽  
Before And After ◽  
Control And Prevention ◽  
Youtube Videos ◽  
Model Training ◽  
Village Government ◽  
And Control ◽  
The Village

Training of health cadres on the prevention and control of Covid-19 is an important and strategic effort. Health cadres who have been trained can participate in screening suspected or suspected cases and can help carry out quarantine or self-isolation or immediately direct them to a hospital if they are suspected of being infected with Covid-19. The purpose of this community service is to increase the knowledge of health cadres related to the control and prevention of Covid-19 in Pujon Kidul Village using an online platform with YouTube videos. good use of media with a maximum combination of modules and videos in providing knowledge to participants. While the weakness is the accuracy and honesty of participants in filling out the questionnaire between before and after the training. Our online training model cannot control the existing situation and conditions, including during the evaluation. Suggestion: online model training needs to be strengthened by the village government by providing online media at the village hall so that it can be accessed by the entire structure of the Covid-19 task force team at the village level.

scholarly journals Highly-transmissible Variants of SARS-CoV-2 May Be More Susceptible to Drug Therapy Than Wild Type Strains

2021 ◽  
Author(s):  
Verena Schöning ◽  
Charlotte Kern ◽  
Carlos Chaccour ◽  
Felix Hammann
Keyword(s):  
Antiviral Therapy ◽  
Antiviral Treatment ◽  
Antiviral Drug ◽  
Drug Regimen ◽  
Wild Type ◽  
Post Hoc Analysis ◽  
Slowing Down ◽  
Post Hoc ◽  
Type Strains

Abstract As of March 2021, no antiviral drug regimen has proved effective against SARS-CoV-2 infection. With the pandemic showing no signs of slowing down, and vaccine campaigns only starting to be rolled out, we appear to have few options other than non-pharmacological measures. Emerging Variants of Concern (VOCs), e.g. B1.1.7, B.1.351, and B.1.1.248, however, are characterized by higher transmissibility (R0). Here we model and simulate the effect of altered R0 on viral load profiles, and its impact on antiviral therapy. As a hypothetical case study, we simulated treatment with ivermectin 600µg/kg for 3 days initiated at different time points around the infection. Simulated mutations range from 1.25 to 2-fold greater infectivity, but also include putative co-adapted variants with lower transmissibility (0.75-fold).Antiviral efficacy was correlated with R0, making highly transmissible VOCs more sensitive to antiviral therapy. Viral exposure was reduced by 42% compared to 22% in wild type if treatment was started on inoculation. Less transmissible variants appear less susceptible.Our findings suggest there may be a role for pre- or post-exposure prophylactic antiviral treatment in areas with presence of highly transmissible variants. Furthermore, clinical trials with borderline efficacious results should consider identifying VOCs and examine their impact in post-hoc analysis.

scholarly journals Efficient Use of GPU Memory for Large-Scale Deep Learning Model Training

2021 ◽  
Vol 11 (21) ◽  
pp. 10377
Author(s):  
Hyeonseong Choi ◽  
Jaehwan Lee
Keyword(s):  
Deep Learning ◽  
Large Scale ◽  
Training Model ◽  
High Accuracy ◽  
Access Pattern ◽  
Learning Models ◽  
Training Models ◽  
Model Training ◽  
Access Patterns ◽  
Deep Learning Model

To achieve high accuracy when performing deep learning, it is necessary to use a large-scale training model. However, due to the limitations of GPU memory, it is difficult to train large-scale training models within a single GPU. NVIDIA introduced a technology called CUDA Unified Memory with CUDA 6 to overcome the limitations of GPU memory by virtually combining GPU memory and CPU memory. In addition, in CUDA 8, memory advise options are introduced to efficiently utilize CUDA Unified Memory. In this work, we propose a newly optimized scheme based on CUDA Unified Memory to efficiently use GPU memory by applying different memory advise to each data type according to access patterns in deep learning training. We apply CUDA Unified Memory technology to PyTorch to see the performance of large-scale learning models through the expanded GPU memory. We conduct comprehensive experiments on how to efficiently utilize Unified Memory by applying memory advises when performing deep learning. As a result, when the data used for deep learning are divided into three types and a memory advise is applied to the data according to the access pattern, the deep learning execution time is reduced by 9.4% compared to the default Unified Memory.

scholarly journals DEVELOPMENT OF FUTSAL BASIC TECHNIQUE TRAINING MODEL FOR BEGINNER BASED ON PLAYING METHODS

2017 ◽  
Vol 3 (1) ◽  
pp. 26
Author(s):  
Febi Kurniawan ◽  
Apta Mylsidayu
Keyword(s):  
Research And Development ◽  
Training Model ◽  
Qualitative Approach ◽  
Training Manual ◽  
Development Method ◽  
Basic Technique ◽  
Model Training

AbstractThe objective of this study is to create a development of futsal basic technique training model for beginner based on valid and reliable play method that is tailored to player’s demand in order to be more convenient for the coach to train futsal by using manual. This research uses quantitative and qualitative approach by using research and development method of Borg and Gall model consisting of 10 stages. The result of this research shows that the product of this development of futsal basic technique training model for beginner based on playing method is nothing to be revised fundamentally, all the indicators have met the standard and feasible to be used, and then the final result on the development of the training manual, so the product developed by researcher can be used as a reference for players or coaches because it can be used as a guidance to increase the effectiveness of futsal basic technique exercise process, besides, the players and coaches are able to obtain various futsal basic technique training based on various playing methods.Keywords: Model, training, basic technique, futsal, beginner player, playing method.

Sign in / Sign up

Export Citation Format

Share Document