http://repo.lib.jfn.ac.lk/ujrr/handle/123456789/110 2026-04-07T17:40:08Z 2026-04-07T17:40:08Z Gowtham, P. Niranjan, M. Kaneswaran, A. http://repo.lib.jfn.ac.lk/ujrr/handle/123456789/10042 2024-01-16T04:43:03Z 2022-01-01T00:00:00Z

Automated gastrointestinal abnormalities detection from endoscopic images Gowtham, P.; Niranjan, M.; Kaneswaran, A. Impressive high performance reported in the use of machine learning on computer vision problems is often due to the availability of very large datasets with which deep neural networks can be trained. With inference from medical images, however, this is not the case and available data is often only a small fraction in size in comparison to benchmark natural scene recognition problems. To circumvent this problem, transfer learning is often applied, where a model trained on a large natural image corpus is adapted, or pre-trained, to model the medical problem. In this work, we consider transfer learning applied to a specific medical diagnostics problem, that of abnormality detection in the gastrointestinal tract of a human body using images obtained during endoscopy. We carry out a search over several image recognition architectures and adapt pretrained models to the endoscopy problem. Using the benchmark KVASIR dataset, we show that transfer learning is effective in outperforming previously reported results, at an accuracy of 98.5±0.27.

2022-01-01T00:00:00Z Jananie, J. Li Chen Fei Xu BoLi http://repo.lib.jfn.ac.lk/ujrr/handle/123456789/10022 2023-12-29T08:01:20Z 2021-01-01T00:00:00Z

Astra: Autonomous Server less Analytics with Cost-Efficiency and QoS-Awareness Jananie, J.; Li Chen; Fei Xu; BoLi Withtheabilitytosimplifythecodedeployment withone-clickuploadandlightweightexecution,serverlesscomputinghasemergedasapromisingparadigmwithincreasing popularity.However,thereremainopenchallengeswhenadapting data-intensiveanalyticsapplicationstotheserverlesscontext,in whichusersofserverlessanalyticsencounterwiththedifficultyin coordinatingcomputationacrossdifferentstagesandprovisioningresourcesinalargeconfigurationspace.Thispaperpresents ourdesignandimplementationofAstra,whichconfiguresand orchestratesserverlessanalyticsjobsinanautonomousmanner, whiletakingintoaccountflexibly-specifieduserrequirements. Astrareliesonthemodelingofperformanceandcostwhich characterizestheintricateinterplayamongmulti-dimensional factors(e.g.,functionmemorysize,degreeofparallelismat eachstage).Weformulateanoptimizationproblembasedon user-specificrequirementstowardsperformanceenhancementor costreduction,anddevelopasetofalgorithmsbasedongraph theorytoobtainoptimaljobexecution.WedeployAstrainthe AWSLambdaplatformandconductreal-worldexperimentsover threerepresentativebenchmarkswithdifferentscales.Results demonstratethatAstracanachievetheoptimalexecutiondecision forserverlessanalytics,byimprovingtheperformanceof21% to60%underagivenbudgetconstraint,andresultingina costreductionof20%to80%withoutviolatingperformance requirement,whencomparedwiththreebaselineconfiguration algorithms.

2021-01-01T00:00:00Z Jananie, J. LiChen FeiXu BoLi http://repo.lib.jfn.ac.lk/ujrr/handle/123456789/10021 2023-12-29T07:19:36Z 2022-01-01T00:00:00Z

Astrea: Auto-Serverless Analytics Towards Cost-Efficiency and QoS-Awareness Jananie, J.; LiChen; FeiXu; BoLi Withtheabilitytosimplifythecodedeploymentwithone-clickuploadandlightweightexecution,serverlesscomputinghas emergedasapromisingparadigmwithincreasingpopularity.However,thereremainopenchallengeswhenadaptingdata-intensive analyticsapplicationstotheserverlesscontext,inwhichusersofserverlessanalyticsencounterthedifficultyincoordinatingcomputation acrossdifferentstagesandprovisioningresourcesinalargeconfigurationspace.Thispaperpresentsourdesignandimplementationof Astrea,whichconfiguresandorchestratesserverlessanalyticsjobsinanautonomousmanner,whiletakingintoaccountflexibly-specified userrequirements.Astreareliesonthemodelingofperformanceandcostwhichcharacterizestheintricateinterplayamongmultidimensionalfactors(e.g.,functionmemorysize,degreeofparallelismateachstage).Weformulateanoptimizationproblembasedon user-specificrequirementstowardsperformanceenhancementorcostreduction,anddevelopasetofalgorithmsbasedongraph theorytoobtaintheoptimaljobexecution.WedeployAstreaintheAWSLambdaplatformandconductreal-worldexperimentsover representativebenchmarks,includingBigDataanalyticsandmachinelearningworkloads,atdifferentscales.Extensiveresults demonstratethatAstreacanachievetheoptimalexecutiondecisionforserverlessdataanalytics,incomparisonwithvariousprovisioning anddeploymentbaselines.Forexample,whencomparedwiththreeprovisioningbaselines,Astreamanagestoreducethejobcompletion timeby21%to69%underagivenbudgetconstraint,whilesavingcostby20%to84%withoutviolatingperformancerequirements.

2022-01-01T00:00:00Z Pratheeba, J. http://repo.lib.jfn.ac.lk/ujrr/handle/123456789/9961 2023-12-20T04:00:38Z 2022-01-01T00:00:00Z

Comprehensive Machine Learning Analysis on the Phenotypes of COVID-19 Patients Using Transcription Data Pratheeba, J. Purpose: Evolving technologies allow us to measure human molecular data in a wide reach. Those data are extensively used by researchers in many studies and help in advancements of medical field. Transcriptome, proteome, metabolome, and epigenome are few such molecular data. This study utilizes the transcriptome data of COVID-19 patients to uncover the dysregulated genes in the SARS-COV-2. Method: Selected genes are used in machine learning models to predict various phenotypes of those patients. Ten different phenotypes are studied here such as time since onset, COVID-19 status, connection between age and COVID-19, hospitalization status and ICU status, using classification models. Further, this study compares molecular characterization of COVID-19 patients with other respiratory diseases. Results: Gene ontology analysis on the selected features shows that they are highly related to viral infection. Features are selected using two methods and selected features are individually used in the classification of patients using six different machine learning algorithms. For each of the selected phenotype, results are compared to find the best prediction model. Conclusion: Even though, there are not any significant differences between the feature selection methods, random forest and SVM performs very well throughout all the phenotype studies

2022-01-01T00:00:00Z