Developing ML solutionsFull notesSummaryRingkasanStoriesPracticeMachine Learning Development Lifecycle10 exam-style questions on this lesson.Question 1 of 10What is the first phase of the ML development lifecycle?AData collectionBBusiness goal identificationCDeploymentDModel trainingCheck answerQuestion 2 of 10Which sequence shows the ML lifecycle phases in the correct order?ABusiness goal identification, model development, data processing, ML problem framing, monitoring, deployment, retrainingBData processing, business goal identification, model development, ML problem framing, deployment, monitoring, retrainingCBusiness goal identification, ML problem framing, data processing, model development, deployment, monitoring, retrainingDML problem framing, data processing, business goal identification, deployment, model development, retraining, monitoringCheck answerQuestion 3 of 10During which phase do stakeholders define the KPIs that will measure success?AModel developmentBMonitoringCML problem framingDBusiness goal identificationCheck answerQuestion 4 of 10Amazon wanted to reduce call transfers by predicting which agent skill each customer call needs. How was this problem framed?AMulticlass classification with supervised learning on historical callsBClustering of historical calls with reinforcement learningCBinary classification of calls with no labelsDRegression on call length with unsupervised learningCheck answerQuestion 5 of 10Which data split is typical for training, validation, and test sets?A100/0/0B80/10/10 or 70/15/15C10/10/80D50/0/50Check answerQuestion 6 of 10Why should a model never be evaluated on its training data?AIt's forbidden by AWSBEvaluation requires labeled data and training data has noneCThe model has already seen it, so the score won't show how it performs on new dataDTraining data is too smallCheck answerQuestion 7 of 10A team sets the learning rate too high. What is the likely result?AThe model becomes perfectly accurateBTraining takes far too long but convergesCThe data split changesDTraining may never convergeCheck answerQuestion 8 of 10What is feature engineering?ACreating and transforming the input variables the model learns fromBMonitoring the model in production for driftCBuying faster hardware to speed up trainingDChoosing the format of the model's predictionsCheck answerQuestion 9 of 10What is hyperparameter optimization?ALearning the model's weights from the training dataBSearching for the best training settings, such as the learning rateCDeploying the trained model to a real-time endpointDRemoving outliers and duplicates from the dataCheck answerQuestion 10 of 10In the Amazon call center example, the team merged all Kindle-related skills into one label. Which phase was this?ADeploymentBBusiness goal identificationCData preprocessingDMonitoringCheck answerPrinciples of Human-Centered Design for Explainable AI8 exam-style questions on this lesson.Developing ML Solutions with Amazon SageMaker AI10 exam-style questions on this lesson.