Machine Learning Flashcards
Types of AWS Machine Learning Solutions
Rekognition - automate your image and video analysis using AI/ML
Comprehend - natural-language processing (NLP) service that finds relationships in text.
Polly - text to speech
SageMaker - build, train, deploy machine learning models
Translate - language translation
Lex - build conversational interfaces
Rekognition*
Rekognition allows you to automate your image and video analysis.
Use cases: content moderation, face detection and analysis,
real world scenario: Rekognition could help Alfredo identify the toppings on his pizzas to make sure they are being made consistently. This would be a great way to train new employees and ensure food quality!
Comprehend
Comprehend is a natural-language processing (NLP)
service that finds relationships in text.
Comprehend in the Real World: Comprehend could help Alfredo process social media posts by looking for words that show customer sentiment about his pizza — words like delicious, tasty, cold, or bad taste. This is a great way to tell if his customers are happy.
Polly
Polly turns text into speech. Can be converted into multiple languages.
Polly in the Real World: Polly could convert the text on a blog post to speech that could then be downloaded or replayed in MP3 format. Audio is often a great complement to written communication.
SageMaker*
SageMaker helps you build, train, and deploy machine learning models quickly.
SageMaker in the Real World: Companies like Netflix and Amazon use machine learning models to recommend movies and products to buy. SageMaker is a great tool for creating these models.
Translate
Translate provides language translation.
Translate in the Real World: Add localization to websites or applications. Translate allows you to add localization to your applications to support your diverse user base. Translate supports several popular languages. Hello, Hola, Ciao, Niheo, konnichiwa
- it is easy to translate, scalable, cost effective and highly accurate
Amazon Forecast
A time-series forecasting services that uses machine learning and is built to give you important business insight
Amazon Fraud Detector
AWS AI service that is built to detect fraud in your data. You can create a fraud detection machine learning model that is based on your data
Amazon Transcribe
Converts speech (video or audio) to text. Can be used to generate subtitles.
Amazon Lex
Allows you to build conversational interfaces I your application using natural language models
SageMaker; Deployment Types
Offline Usage: Asynchronous or bath usage. Used when you need immediate response.
Online Usage: Synchronous or real time usage. Used when you don’t need immediately
SageMaker: Components
Notebook - Access a managed Jupyter notebook environment. Where you write machine learning models.
Training - Train and tune models
Inference - Package and deploy your machine
leaning models at scale
SageMaker: Creation Flow
1st - Create a model
2nd - Create an endpoint configuration
3rd - Create an Endpoint.
SageMaker Neo
- Customize your machine learning models for specific CPU hardware, such as ARM, Inter, and NVIDIA processors.
- It includes a complier to convert the machine learning model to an environment that is optimized to excecute the model on the target architecture.
Elastic infrence
- you can decrease the cost of running Sagemaker by deploying Elastic Inference (EI).
- EI speeds up throughput and decreases latency of real-time inferences deployed on Sagemaker hosted services using only CPU-bases instances.
