Transcription
Machine Learning SimplifiedGraham Dudgeon, PhDPrincipal Industry Manager 2015 The MathWorks, Inc.1
2
Consider Machine Learning WhenBecause algorithms canSolution is too complex for hand written rules or equationslearn complex nonlinear relationshipsSpeech RecognitionObject RecognitionEngine Health MonitoringSolution needs to adapt with changing dataupdate as more databecomes availableWeather ForecastingEnergy Load ForecastingStock Market PredictionSolution needs to scalelearn efficiently fromvery large data setsIoT AnalyticsTaxi AvailabilityAirline Flight Delays3
What is Machine Learning?Machine learning algorithms use computational methods to “learn” informationdirectly from data without assuming a predetermined equation as a modelTraining dataExtract FeaturesTrain Models DogCarCatBirdDogs4
Challenges80% effortAccessDataExtractFeaturesTimeconsuming fornon-datascience expertsDevelopModelsRequires handcoding,programmingskillsShareModels5
Challenges from our Customers Convert unreadable data into a usable format. Automate filtering, spectral analysis, andtransform steps for multiple trucks and regions. Develop a predictive maintenance system toreduce pump equipment costs and downtime. Lack of experience with neural networks ormachine learning. Develop a prototype quickly, relying on functionsthat have been deployed across ASML’s large,diverse user base and maintained by dedicatedprofessionals.6
New MATLAB framework makes machine learningeasy and accessible for Engineers7
MATLAB makes Machine Learning Easy and Accessible AccessPreprocessDevelop ModelsShare, Integrate with industry enabling non- from ideaproven solutionsexpertsto product8
MATLAB makes Machine Learning Easy and Accessible AccessPreprocessDevelop ModelsShare, Integrate with industry enabling non- from ideaproven solutionsexpertsto product9
Using Machine Learningto build and deploy a predictive maintenance system1TBAnalytics andMachine Learningplus signal processing,neural networks & morePump logsof temperature, pressure& other dataPredictive Modeldeployed to drill siteMaintenanceNeeded10
Using Machine Learningto build and deploy a predictive maintenance system1TBAnalytics andMachine Learningplus signal processing,neural networks & morePump logsof temperature, pressure& other dataPredictive Modeldeployed to drill siteMaintenanceNeeded11
12
AI, Machine Learning and Deep LearningArtificial IntelligenceReasoningApplication BreadthPerceptionKnowledge RepresentationMachine TranslationMachine LearningWeather ForecastingSpam DetectionRecommender SystemsComputer Board GamesMedical DiagnosisExpert Systems1950sHealth Monitoring1980sTimelineAlgorithmic TradingDeep LearningFraud DetectionBioinformaticsInteractive ProgramsSentiment AnalysisAutomated DrivingObject RecognitionRoboticsSpeech RecognitionToday13
What is Deep Learning?CatDeep learning is a type of machine learning that learns tasksdirectly from dataDogLearned Features CarDogCatBirdBirdCar14
Why is Deep Learning So Popular Now?HumanAccuracySource: ILSVRC Top-5 Error on ImageNet15
Deep Learning EnablersAcceleration with GPUsMassive sets of labeled dataAvailability of state of the art models from experts16
New MATLAB framework makes deep learningeasy and accessible for Engineers17
MATLAB makes Deep Learning Easy and Accessible Handle large images sets Accelerate with GPUs Visualize and debug networks Access pre-trained models18
Deep Learning is Changing the WorldTrain from scratch!Transferlearning in 10lines of code!19
Deep Learning is changing the worldTrain from scratch!Transferlearning in 10lines of code!20
Our Customers Achievements“MATLAB gave us the ability to convert previously unreadabledata into a usable format; automate filtering, spectral analysis, andtransform steps for multiple trucks and regions; and ultimately, applymachine learning techniques in real time to predict the idealtime to perform maintenance.”Gulshan SinghBaker Hughes“As a process engineer I had no experience with neural networks ormachine learning. I worked through the MATLAB examples to find the bestmachine learning functions for generating virtual metrology. I couldn’thave done this in C or Python—it would’ve taken too long tofind, validate, and integrate the right packages.”Emil Schmitt-WeaverASML21
Summary of results Savings of more than 10 million projected Development time reduced tenfold Multiple types of data easily accessed Industry leadership established Potential manufacturing improvements identified Maintenance overhead minimized22
How to get started? Data Processing Machine Learning Computer VisionPublicOn-Site23
How to get started? Data Processing Machine Learning Computer VisionPublicOn-Site24
MATLABData AnalyticsData Processing andVisualizationStatisticsMachine LearningOptimization TechniquesParallel ComputingControl System DesignSignal ProcessingCommunication SystemsLTE SystemsComputationalFinanceProgramming TechniquesBuilding InteractiveApplicationsObject-Oriented ProgrammingRisk ManagementTime-Series ModellingSIMULINKModel-BasedDesignImplementing MBD WorkflowModel Management andArchitectureVerification and ValidationSTATEFLOWCode GenerationRapid Prototyping and HILSimulationEmbedded SystemsFPGA DesignGenerating HDL CodeXilinx Zynq SoCsAUTOSAR Event-Based ModelingCode IntegrationTMSignal ProcessingMATLAB CoderInterfacing with C-code ApplicationSpecificApplicationDevelopmentCode Generation Using MATLABUsing SimulinkImage and VideoProcessingImage ProcessingComputer VisionIntegrating C and MATLABSimscapeGeneral SimscapeTMSimscape MultibodyTMSimscape DrivelimeTMSimscape FluidsTMSimscape Power SystemsTM PolyspacePolyspace Code .html25
machine learning techniques in real time to predict. the ideal time to perform maintenance." Gulshan Singh Baker Hughes "As a process engineer I had no experience with neural networks or machine learning. I worked through the MATLAB examples to find the best machine learning functions for generating virtual metrology. I couldn't