AWS Machine Learning Specialty

AWS Machine Learning Specialty

AWS SaaS

Check out this list before spending precious company resources training your own model. Trust me, they are far superiour.

• Amazon Forecast - resouce forecasting from various datasources
• Amazon Lookout for Metrics - anomaly detection for cloudwatch metrics
• Amazon Fraud Detector
• Amazon Personalize - personal recommendation engine
• Amazon Polly - text to speech (with human-like qualities)
• Amazon Transcribe - speech to text
• Amazon Translate - language translation
• Amazon Kendra - AI-powered search
• Amazon Comprehend - natural language processing (keyphrase extraction, sentiment analysis, entity recognition)
• Amazon Rekognition - image entity detection
• Amazon Textract - OCR, form and table data extraction
• Amazon Lex - chatbot

Development Platform

So, what you are trying to build is a novel idea... i guess we have no choice but to train your own model. At this point, please let AWS help you not to worry about the end-to-end process of testing and deployment.

• SageMaker Data Studio - Jupyter on steroids
• SageMaker Feature Store - offline and online storage for curated features
• SageMaker Data Wrangler - transforms and analyze data
• SageMaker Clarify - offers explainability and bias detection

Hardware

• AWS DeepLens (AI-Powered Camera)
• Amazon Elastic Inference - on-demand GPU power

Targetting the first ML Project in your Organization

ML Lifecylce

• Addressing Overfitting(high variance) /Underfitting (high bias)
  • Regularization - minimize overfitting, add cost to to the parameter L1 (absolute value of the sum), L2 (square value of the parameters)
  • Feature extraction - create new features from existing features
• Tuning
  • Hyperparameter Tuning - depends on model architecture, set by the engineer, not part of the estimation
• Model Evaluation Metrics
  • Accuracy
  • Precision
  • Recall
  • F1