AWS Machine Learning Specialty Practice 2021

Realistic practice exam based on the most recent AWS Certified Machine Learning Specialty exam.

• Just like the actual exam this practice test has

  • Test 1: 30 questions

  • Test 2: 65 questions and takes 170 minutes

    
    

• Questions are mapped based on the actual exam domains:

  • Data Engineering

  • Exploratory Data Analysis

  • Modeling

  • Machine Learning Implementation and Operations.

• Suggested background knowledge

  • Data Engineering

    • AWS services:

      • Glue, EMR ( Apache Spark, Hive metastore),  Athena

      • Kinesis family (Streams, data analytics, firehose, video streams)

      • S3, QuickSight

      • Data/File formats (Avro, Parquet, CSV, protobuf recordIO)

  • Exploratory Data Analysis

    • Handling missing values (Imputation: median, mean, most frequent, using ML model)

    • Feature scaling

    • Feature engineering

    • Handling outliers

    • One-hot encoding

    • Binning

    • Text preprocessing

  • Modeling

    • supervised machine learning ( Classification and Regression Algorithms)

    • unsupervised machine learning ( K-Means clustering, PCA)

    • Hyperparameter tuning ( supervised machine learning, deep learning)

    • Performance metrics ( accuracy, RMSE, F1 score, AUC, ROC, Precision, Recall)

    • Tuning deep learning networks ( how to prevent overfitting)

    • AWS ML services: Lex, Polly, Transcribe, Translate, Comprehend

    • SageMaker built-in algorithms: BlazingText, Object2Vec, DeepAR, LDA, Linear Learner, etc.

  • ML Implementation and Operations

    • Amazon SageMaker train and deploy a model

      • Inference pipeline, batch transform, inference endpoints, production variants, hosting services

    • Amazon SageMaker security (data encryption at rest and in transit)

    • Distributed training on Amazon SageMaker ( Using GPUs)

    • AWS SageMaker roles

    • Bring your own model container ( e.g. developed using scikit-learn)

    • Customize SageMaker built-in algorithm containers

    • How to develop and deploy deep learning models on frameworks such as Tensoflow, MXNeT