Classical Scheduling Research Articles

Use of An Ophthalmology Tutorial to Improve Resident Comfort with the Emergency Eye Exam.

This tutorial should be utilized for emergency medicine (EM) interns and junior residents. Ophthalmology is characteristically a weak area in both medical school and resident education. Medical students are rarely given formal didactic education on the use of the slit lamp or a systematic approach to examining the eye. For EM residents, this leads to inefficient and uncomfortable encounters with patients with eye complaints. We sought to develop a comprehensive emergency ophthalmology tutorial utilizing asynchronous learning followed by a hands-on skill session that would address this need. By the end of this small group didactic, learners will be able to: 1) demonstrate ability to focus on the various components of the slit lamp exam 2) demonstrate understanding of a systematic approach to the eye exam 3) demonstrate appropriate use of the Diaton, iCare, and Tonopen tonometers. This two-hour small group didactic combines hands-on learning sessions to learn the slit lamp exam and tonometry measurement, with a systematic review of the eye exam to help learners better organize their exams and understand the use of necessary tools. The emergency ophthalmology tutorial was initially designed as an education project in which we collected pre- and post-participation surveys regarding resident comfort with various components of the emergency eye exam. After the course residents received a post-course survey to complete. Given the positive feedback we received from our residents regarding the tutorial, we applied for Institutional Review Board (IRB) approval to publish our retrospective survey data. Our IRB waived the need for participant consent. Twelve emergency medicine residents including 11 interns and one post-graduate year (PGY) 2 resident participated in the emergency ophthalmology tutorial as part of our intern boot camp in July of 2021. Twelve PGY-1 residents initially signed up for the course and filled out the pre-participation survey but one of them was not able to attend their scheduled class, so a PGY-2 resident requested to attend.Prior to the course, we used a Likert scale from 1-7, finding that 61.5% (8/13) of participants felt very uncomfortable with performing slit lamp exams, 84.6% (11/13) felt very uncomfortable with using the Diaton tonometer, 76.9% (10/13) felt very uncomfortable with using the iCare tonometer, and 69.3% (9/13) felt uncomfortable or very uncomfortable with using a systematic approach to examining the eye. After the course, 75% (9/12) of participants felt that the course exceeded expectations in ensuring their ability to perform the subcomponents of the slit lamp exam, 75% (9/12) and 83.3% (10/12) of participants felt that the course exceeded expectations in ensuring their ability to use the Diaton and iCare tonometers, respectively, and 91.7% (11/12) felt that the course exceeded expectations in ensuring their ability to perform a systematic eye exam. Participation in a 2-hour emergency ophthalmology tutorial with assigned asynchronous pre-course work improved emergency medicine resident comfort with various components of the eye exam. Emergency ophthalmology, eye exam, slit lamp, tonometry.

Cloneable Jellyfish Search Optimizer Based Task Scheduling in Cloud Environments

For cloud environments, task scheduling focusing on the optimal completion time (makespan) is vital. Metaheuristic approaches can be used to produce efficient solutions that will provide important cost savings to both the cloud service provider and the clients. On the other hand, since there is a high probability of getting stuck in local minima in metaheuristic solutions due to the type of problem, it may not always be possible to quickly reach the optimal solution. This study, using a metaheuristic approach, proposes a solution based on the Cloneable Jellyfish Algorithm for optimal task distribution in cloud environments. The unique feature of the proposed algorithm is that it allows dynamic population growth to be carried out in a controlled manner in order not to get stuck in local minima during the exploration phase. In addition, this algorithm, which uses a different cloning mechanism so that similar candidates are not generated in the population growth, has made it possible to achieve the optimal solution in a shorter time. To observe the solution performance, cloud environment simulations created in the Cloudsim simulator have been used. In experiments, the success of the proposed solution compared to classical scheduling algorithms has been proven.

Emergency Remote Teaching Based on Education 4.0: an Experience Report about the Methodology adopted in the Experimental HCI Class

In Education 4.0, the professor is challenged to teach more than curriculum content, prompting the development of 21st-Century competencies and skills. Thus, this paper presents the Emergency Remote Teaching (ERT) methodology used in the Experimental Human-Computer Interaction (HCI) classes during the COVID-19 pandemic, based on Education 4.0. This class was taught to final year students of the undergraduate courses in Computer Science and Biomedical Informatics and graduate students in Informatics at a University. The teaching methodology adopted in this discipline sought to develop skills considered relevant for a researcher, such as recognizing research problems and learning new processes. The classes were guided by the Flex learning model, which consisted of synchronous and asynchronous classes. Thus, the schedule of classes needed to be adapted, including collaborative and diversified activities. We collected the student’s opinions about the teaching methodology and the skills developed in the class and analyzed the results qualitatively. The results showed the difficulties and problems faced by students with the resources and materials in ERT, such as students having internet connectivity problems. However, the teacher made the recorded classes available, and this problem was overcome. The results also show that the deficit of skills, such as autonomy, initiative, and independence, hampered students' performance in ERT. The findings can help other teachers to improve the remote teaching of this class and indicate the possibilities of ERT in the Education 4.0.

Clemson University’s Rotational Attendance Plan During COVID-19

The COVID-19 pandemic forced universities to upend their class scheduling. At Clemson University, the administration implemented hybrid schedules for fall 2020, in which students attend classes partly online and partly in person. To limit exposures of COVID-19 in the classroom, we propose two rotational attendance models (the three-cohort model and the once-a-week model) that aim to allow in-person classroom time and minimize exposure between students. In a baseline strategy, students would interact with an average of 84 students per week and attend class in person 2.6 days a week. By contrast, the three-cohort model and once-a-week model achieve about 57 and 83 student interactions per week and 1.6 and 1.9 in-person student attendance days a week, respectively. Although these figures of merit may imply that the three-cohort model is preferable, it achieves its results by forcing about 1,600 of the 21,000 students who want to attend courses in person to participate online instead and forcing courses that meet twice a week to be attended twice in a three-week rotation. Considering the tradeoffs between the figures of merit related to student interaction and anticipated implementation challenges, Clemson University implemented the once-a-week model for fall 2020 and spring 2021. History: This paper was refereed. This article has been selected for inclusion in the Special Issue on Analytics Remedies to COVID-19.

Effects of a blended classroom-based intervention on aerobic fitness, motor skills, inhibition, and daytime sleepiness among Hong Kong children.

In response to the all-round development of primary school children, this study aimed to evaluate the effectiveness of a blended intervention program on children's health-related outcomes of aerobic fitness, motor skills, inhibition and daytime sleepiness in classroom settings. Three experimental conditions include: (1) the “Stand + Move” group combining sit-stand desks and physical activity (PA) recess, (2) “Move” group with PA recess only, and (3) Control group (CG; normal class schedule). A total of 64 primary school children (37.5% girls and 62.5% boys, M [SD] = 9.6 [0.61], BMI mean = 17.0 ± 3.0) participated in all assessments, including aerobic fitness, motor skills, inhibitory control, and daytime sleepiness. The baseline data collection starts from January 2019, with the intervention lasting for 13 weeks and followed by post-intervention and follow up tests conducted in July and October 2019. There was a significant interaction effect on aerobic fitness [F(2,76) = 10.62, p < 0.001, η2 = 0.22] after the intervention period, whereas no significant interaction was observed for other variables. Significant main effects were observed in aerobic fitness (pre – post: −11.75 and −7.22) for both experimental groups, with the blended group showing greater improvements immediately post the test, while motor skills only showed a significant increase at the three-month follow-up, with the greatest increase in the blended group (pre-follow-up: −2.50). For inhibition control and daytime sleepiness, better improvements were shown for the experimental groups than for the control group. The blended designed intervention, by incorporating multiple components as an innovative strategy to reconstruct children's traditional classroom environment in Hong Kong, has demonstrated improved physical and psychological development of school children.Clinical trial registrationChiCTR, ChiCTR2000035038. Registered 29 July 2020 – Retrospectively registered. http://www.chictr.org.cn/hvshowproject.aspx?id=46038.

Problems of Online Learning according to Students during the Covid-19 Pandemic

This study aims at describing the types of problems experienced by students while participating in online learning, online learning media that students prefer, and evaluation of the implementation of online learning according to students at the Indonesian Muslim University during the Covid-19 pandemic. This study uses a survey method. The research sample is the students of the Faculty of Letters. The data collection technique used in this study is Google form questionnaire and analyzed by using quantitative descriptive statistics. The results show that the types of problems experienced by students while participating in online learning are generally caused by two factors, namely internal factors: unable to focus on attending lectures, problem on internet network, lack of cost to buy internet quota, and external factors: class schedules change frequently, the use of less effective learning media, redundant assignments. Online learning media that are generally preferred by students are Zoom Meeting, UMI Kalam e-Learning, and WhatsApp. According to students, some points that need to be evaluated in online learning include: lecturers often change class schedules and immediately organize face-to-face learning, lecturer learning strategies are less effective, giving excessive assignments, internet connections often in problems, students need to receive internet quota fund, and lecturers are still less skill in using digital learning media.

Conceptualizing “classroom teaching-learning process” that engages students in Indian business school

PurposeCase-based classroom teaching-learning process (hereafter, case method) has provided a very productive teaching-learning environment for a long time. In the case method, students are expected to meet some prerequisites, such as reading and analyzing the case in advance, listening to the classroom discussion and actively participating in the discussion. However, it is frequently reported in Indian business schools that students do not prepare the assigned case before the scheduled class. The under-preparation of cases results in low-quality discussion, high absenteeism, passive attitude and lack of energy.Design/methodology/approachThis study modeled the case method using an IGEO (input-guide-enablers-output, commonly used in any process modeling) framework to identify challenges in the case-based classroom teaching-learning process. A novel customized classroom teaching-learning process called the EPDE (explain, practice, discuss, explore) method replaced the case method. These two teaching-learning processes were used for teaching two groups of MBA students.FindingsThe novel EPDE method effectively addressed the case method challenges. It resulted in better learning outcomes in the Indian B-school context.Originality/valueThe purpose of the study is to examine the impact of an alternative to the traditional case study method in a college classroom among MBA students. Two points make this study original and unique: (1) The IGOE process modeling framework is used to model teaching-learning processes, such as the case and EPDE methods. Using IGOE for teaching-learning processes is unique and is not available in the literature and (2) the EPDE method is a novel concept.

Application of an Environmental, Social and Governance (ESG) Course to Enhance the Knowledge of Sustainable Finance of International Chinese College Students in Thailand

The aim of this study was to build an ESG course for integration into a Financial Statement Analysis course to develop the knowledge of sustainable finance of international Chinese college students in Thailand. The study was based on a quasi-experimental design and a total of 108 international Chinese students majoring in finance in Thailand participated in the research. 58 of them were assigned to the experimental group, and the remaining 50 to the control group. The intervention of the ESG course was applied to the experimental group for a 45-hour’ period over 6 weeks, while the control group received traditional teaching in the same period of the class schedule. At the same time, an ESG learning assessment was applied to a pre-test and post-test and the data was subjected to a statistical analysis. The results of a one￾way ANCOVA showed that the performance of the experimental group on the ESG learning assessment was better than that of the control group.

Implementation of Back Tracking Algorithm in The Scheduling of Mathematics Study Program Faculty of MIPA Unsoed

The complicated case of scheduling courses at the Mathematics and Natural Sciences Faculty of Mathematics Study Program, Universitas Jenderal Soedirman, is a topic that is quite interesting to discuss and find a solution to using a mathematical method. In fact, manual course scheduling without a method is prone to scheduling errors such as class schedule clashes, clashes in the use of lecture rooms and so on, so a more efficient method of scheduling courses is needed. Scheduling lectures with the backtracking algorithm is a systematic and efficient method of scheduling lectures with influencing factors such as the number of courses, number of sessions, number of rooms, and lecture time. Algorithm backtracking is an algorithm based on Depth First Search to find solutions to problems more efficiently. The back tracking algorithm performs a systematic search for solutions to all possible solutions at each node based on recursive Depth First Search. Depth First Search is a search method that is carried out at one node in each level from the far left. If a solution has not been found, then the search continues on the right hand node. And so on until a solution is found or if you find a dead end it will backtrack to the previous position. If a solution is found, the search will stop even if there are nodes that have not been traced. The implementation of the backtracking algorithm pays close attention to the factors that become obstacles in scheduling courses. The course schedule function with the backtracking algorithm can meet every influencing factor such as the number of courses, rooms, classes, and lecture time so that scheduling lectures with this method is very helpful because the method used is more efficient and can avoid errors in scheduling.

Performant, Multi-Objective Scheduling of Highly Interleaved Task Graphs on Heterogeneous System on Chip Devices

Performance-, power-, and energy-aware scheduling techniques play an essential role in optimally utilizing processing elements (PEs) of heterogeneous systems. List schedulers, a class of low-complexity static schedulers, have commonly been used in static execution scenarios. However, list schedulers are not suitable for runtime decision making, particularly when multiple concurrent applications are interleaved dynamically. For such cases, the static task execution times and expectation of idle PEs assumed by list schedulers lead to inefficient system utilization and poor performance. To address this problem, we present techniques for optimizing execution of list scheduling algorithms in dynamic runtime scenarios via a family of algorithms inspired by the well-known heterogeneous earliest finish time (HEFT) list scheduler. Through dynamically arriving, realistic workload scenarios that are simulated in an open-source discrete event heterogeneous SoC simulator, we exhaustively evaluate each of the proposed algorithms across two SoCs modeled after the Xilinx Zynq Ultrascale+ ZCU102 and O-Droid XU3 development boards. Altogether, depending on the chosen variant in this family of algorithms, we are able to achieve an up to 39% execution time improvement, up to 7.24x algorithmic speedup, or up to 30% energy consumption improvement compared to the baseline HEFT implementation.

The Effects Of A Regular Yoga Practice On Mental Health In College Students During COVID-19

College students often experience stress and anxiety, with many reporting increased stress and anxiety due to the COVID-19 pandemic. A regular yoga practice has been shown to reduce anxiety in this population prior to the COVID-19 pandemic. PURPOSE: The purpose of this study was to determine if a 12-week regular yoga practice taught over Zoom would improve measures of mental health in college students during the COVID-19 pandemic. METHODS: Students were recruited from Yoga 1 classes (Yoga, Y) or from various non-yoga professors at the University (Control, C) at the start of the semester. Upon enrollment, all participants responded to an online survey regarding their level of physical activity (IPAQ short form), perceived daily stress (1 = low; 5 = high, often difficult to handle), and state and trait anxiety (STAI forms Y1 and Y2). During the next 12 weeks, Y students participated in 150 minutes/week of a regular beginner’s yoga practice taught over Zoom. C students participated in their classes as usual and were asked not to make any major changes to their fitness routine. At the completion of Week 12, participants completed the follow up survey, identical to baseline. Due to the small sample size and violation of normality in some variables, all variables were compared via Mann Whitney Rank Sum tests. RESULTS: 24 yoga participants (Y: 17F; 21 ± 1 yrs) and 21 non-yoga participants (C: 18F; 21 ± 1 yrs) completed the entire study with Y students attending at least 80% of their scheduled yoga classes. There were no differences between groups in total MET minutes per week (MEAN ± SEM: Y 4032 ± 611 vs C 4255 ± 728 min, p = 0.979), ratings of daily stress (Y 2.9 ± 0.2 vs C 2.9 ± 0.2, p = 0.761), and STAI Y1 (Y 43.8 ± 2.3 vs C 44.6 ± 2.5, p = 0.900) or Y2 surveys (Y 44.1 ± 2.2 vs C 44.8 ± 3.0, p = 0.838) at baseline. There was also no difference between groups in total MET minutes per week (Y 3905 ± 582 vs C 4233 ± 827 min, p = 0.973), daily stress (Y 2.8 ± 0.2 vs C 2.8 ± 0.2, p = 0.787) STAI Y1 (Y 42.0 ± 1.9 vs C 41.8 ± 2.5, p = 0.750) or Y2 (Y 41.7 ± 2.1 vs C 44.8 ± 2.5 p = 0.387) surveys after 12 weeks. CONCLUSION: 12 weeks of a regular yoga practice taught over Zoom did not improve measures of mental health in college students during the COVID-19 pandemic. These results are not consistent with previous work that suggested the beneficial effects of regular yoga practice on mental health.

An approximate evaluation method for neighbourhood solutions in job shop scheduling problem

Job shop scheduling problem is a classical scheduling problem, and it is very difficult to work out. To solve it well, the meta-heuristic algorithm is a good choice, and the evaluation method of neighbourhood solutions, which affects the efficiency of the algorithm and the quality of the solution, is one of the keys in the algorithm. We propose an approximate evaluation method by exploring domain knowledge in neighbourhood solutions. Firstly, we reduce the computational time of the evaluation by analysing the unnecessary computational operations. Secondly, according to the domain knowledge, we prove that the evaluated value of the neighbourhood solution is the exact value under certain conditions. At the same time, a set of critical parameters are calculated to correct the estimated value of the neighbourhood solutions that do not meet the conditions to improve the evaluation accuracy. With all of these, an approximate evaluation method for neighbourhood solutions in job shop scheduling problems is proposed. The experiments on different numerical instances show the superiority of the method proposed.

Classification and Performance Study of Task Scheduling Algorithms in Cloud Computing Environment

Cloud computing is becoming very common in recent years and is growing rapidly due to its attractive benefits and features such as resource pooling, accessibility, availability, scalability, reliability, cost saving, security, flexibility, on-demand services, pay-per-use services, use from anywhere, quality of service, resilience, etc. With this rapid growth of cloud computing, there may exist too many users that require services or need to execute their tasks simultaneously by resources provided by service providers. To get these services with the best performance, and minimum cost, response time, makespan, effective use of resources, etc. an intelligent and efficient task scheduling technique is required and considered as one of the main and essential issues in the cloud computing environment. It is necessary for allocating tasks to the proper cloud resources and optimizing the overall system performance. To this end, researchers put huge efforts to develop several classes of scheduling algorithms to be suitable for the various computing environments and to satisfy the needs of the various types of individuals and organizations. This research article provides a classification of proposed scheduling strategies and developed algorithms in cloud computing environment along with the evaluation of their performance. A comparison of the performance of these algorithms with existing ones is also given. Additionally, the future research work in the reviewed articles (if available) is also pointed out. This research work includes a review of 88 task scheduling algorithms in cloud computing environment distributed over the seven scheduling classes suggested in this study. Each article deals with a novel scheduling technique and the performance improvement it introduces compared with previously existing task scheduling algorithms. Keywords: Cloud computing, Task scheduling, Load balancing, Makespan, Energy-aware, Turnaround time, Response time, Cost of task, QoS, Multi-objective. DOI: 10.7176/IKM/12-5-03 Publication date: September 30 th 2022

APLIKASI MOBILE ASSISTANT JURUSAN TEKNIK INFORMATIKA UNIVERSITAS PALANGKA RAYA

The process of attendance information that already exists at Palangka Raya University Informatics Engineering cannot be monitored or is still sharing manually using social media, especially such as informing the presence / presence of lecturers on campus for consultation or signature documents. The purpose of this research is to design and build a Mobile Assistant Application for Informatics Engineering at Palangka Raya University, whose development is based on Location Based Service using a prototype methodology with stages of Communication, Quick plan, Modeling Quick Design, Prototype Formation, Deployment Delivery &amp; Feedback. The modeling of the system uses Java Android object-oriented programming and structured for web applications using the programming language using PHP database using MySQL. From the research conducted, the writer implements the results of the research into the Mobile Assistant Application of the Informatics Engineering Department of Palangka Raya University where there are features of lecturer attendance, class schedules and practicum, as well as assignments. The application of Location Based Service in this application is applied to the Google Map API with geofencing, the central point of the Informatics Engineering department is made a marker that generates a radius where if the application user enters into each radius it will produce a room information notification and send an update of the data to the database as the presence status lecturer.

Student Motivation to Follow the Student Creativity Program

This study aims to determine (1) the factors that cause USK FKIP students to be less motivated to participate in PKM and (2) how to increase the motivation of USK FKIP students to participate in PKM. This article uses a qualitative approach with a descriptive type of research. The sampling technique used was quota sampling and data collection techniques in the form of observation, interviews, documentation and data triangulation. The data analysis technique used is data reduction, data presentation, and drawing conclusions. The findings in this study could be presented as follows: The factors causing students to be less motivated to participate in PKM activities are (1) an intrinsic factors (intelligence 25%, attention 19%, interest 14%, motive 12%, maturity 13% and fatigue 17%), (2) the extrinsic factors (class schedules 20%, curriculum 24%, relations between students and students 28%, relations between students and lecturers 28%). Efforts made to increase students motivation to participate in PKM activities are: (1) efforts made by the Vice Dean for student affairs, which consist of intrinsic factors (attention 37.5% and motivation 62.5%) and 100% factors of extrinsic is student-teacher relationship factor (2) efforts made by the study program are intrinsic factors consisting of 22% attention, 37% motivation and 41% readiness and extrinsic factors consist of 27% curriculum and 73% relationship between lecturers and students. The effort expected by students consists of intrinsic factors (intelligence 34% and attention 66%), while extrinsic factors (curriculum 16%, relationships between students and students 40% and relationships between lecturers and students 44%).

Time-efficient physical activity intervention for older adolescents with disability: rationale and study protocol for the Burn 2 Learn adapted (B2La) cluster randomised controlled trial

IntroductionPhysical activity declines during adolescence, with the lowest levels of activity observed among those with disability. Schools are ideal settings to address this issue; however, few school-based interventions have been...

THE EFFECT OF STATISTICS AND PROBABILITY LEARNING MODEL IMPROVEMENT ON STUDENT LEARNING OUTCOMES

The Statistics dan Probability class is one of the elective courses that have to be enrolled by students who have just joined the beginning of the semester. That is because studying Statistics and Probability is crucial in providing the first foundation for knowledge in other scientific domains. In actuality, though, a lot of students struggle with this course. Many students claim that they have trouble understanding the subjects they are learning, and they also have trouble figuring out how to answer the lecturer's questions. The low cognitive ability of students, which results in limited student creativity in problem-solving skills, is one of the elements contributing to students' failure to comprehend the subject. Of course, this is a severe issue if nothing is done right away. Based on these issues, researchers who were also lecturers improved the teaching strategies for all classes. This study aims to determine whether there is substantial interaction between class schedule and the development of learning model that impact exam scores by raising students' Statistics and Probability scores. The two-way Anova analysis principle is the basis of the study methodology. The enhancement of learning model used outside and inside the classroom was found to have a substantial interaction based on the research findings on fifty samples of students selected from different class. A significance score of 0.000 indicates that these findings are significant. In other words, a model is required to achieve better learning outcomes, which might make it simpler for students to commence learning and provide a common purpose for learning. Since multiple class schedules exist, each learning model necessitates a unique management system and learning environment.

Application of the Waterfall Method in Software Design on Android-Based Programming Language Course Applications

In this era of technology and information proliferating, programming skills are needed. Technology affects every area of life from industry, business, communication, transportation, health, and others. Everyone has the same opportunity to learn and master technology thus programming language courses are needed to provide education, innovation, and improvement of skills and abilities in the field of programming and data science to the public. Until now, programming language courses still use the conventional system, where everything is done manually, from class registration, class scheduling, teaching and learning process, and payment processing which results in many archives that must be stored for administrative purposes and require a relatively large amount of time for customers to come to the course location. Therefore, an information technology-based system is needed to fix the weaknesses of the old system. In this study, an Android-based programming language course application is designed to facilitate customers and course owners in teaching and learning activities and transactions. The design in this study uses the waterfall method, which consists of five stages, needs analysis, design, code, testing, and maintenance. The results obtained from testing applications using questionnaires on programming language course applications are 80% stated by course customers, where the application is easy to use, faster, and more practical in registering. In conclusion, this designed application can make it easier for customers to carry out teaching and learning activities and transact quickly and practically.

PEMBELAJARAN MAHASISWA KEPERAWATAN STIKES MUHAMMADIYAH GOMBONG SELAMA PANDEMI COVID-19

Efforts to prevent the spread of Covid-19 were carried out by means of social distancing and self-isolation. This effort has also forced the closure of all schools, both elementary and tertiary levels. School closures were carried out in more than 100 countries. Learning during this pandemic requires STIKes Muhammadiyah Gombong to conduct online or online learning. The purpose of scientific writing is to provide an overview of the learning of nursing students at STIKes Muhammadiyah Gombong during the Covid 19 pandemic. The method used in this study is descriptive by distributing questionnaires and interviews to assess the learning process online. The Results showed that the percentage of student who Disliked online learning is 52.8%, Online learning schedule according to predefined class schedules 48.3%, Liked learning media used by online lecturers 44%, opening E Learning to better master the material 61.4%, less supportive facilities to access online learning media 40.1%, the media used by lecturers was not easily accessible and used 50.3%, online learning did not increase my learning outcomes 57.7%, difficulty understanding material in online learning was 53.4%. Recommendations are needed online learning guides and supporting facilities for the effectiveness of online learning

Online scheduling algorithms for unbiased distributed learning over wireless edge networks

To realize smart IoT services, such as intelligent video surveillance, smart city and autonomous driving, tremendous amount of distributed machine learning jobs will train unbiased models over large datasets collected by geo-distributed wireless edge network, adopting a parameter server (PS) architecture. The training of unbiased distributed learning (UDL) relies on geo-distributed data and brings high response latency and bandwidth consumption, introducing a new challenge: how to schedule UDL jobs such that the response latency (training time) is minimized, meanwhile reducing the expensive bandwidth cost among geo-distributed sites in the edge wireless network. To address it, we propose two online scheduling algorithms, Okita and Okita∗, to achieve long-term overall cost minimization. Okita schedules UDL jobs at each time slot in a preemptive manner to jointly decide the execution time window, the amount of training data, the number and the location of concurrent workers and PSs in each site, whereas Okita∗ schedules jobs via a non-preemptive fashion. To evaluate the practical effectiveness of the proposed algorithms, we implement both testbed experiments and large-scaled simulations. We show that our proposed algorithms can reduce up to 70% of the total training cost, compared to three classical schedulers in today’s cloud system.