Google Cloud Professional Machine Learning Engineer Certification Flashcards¶
26 cards from this guide's fact sheet and notes. Click a question to reveal the answer.
To study with spaced repetition, download flashcards.csv and import it into Anki or Quizlet (comma separated, first row is the header). Back to the study guide.
Exam logistics¶
How long is the Google Cloud Professional Machine Learning Engineer Certification exam?
120 minutes (2 hours)
How many questions are on the Google Cloud Professional Machine Learning Engineer Certification exam?
50-60 questions
What is the passing score for Google Cloud Professional Machine Learning Engineer Certification?
~70% (not officially published)
What does the Google Cloud Professional Machine Learning Engineer Certification exam cost?
$200 USD
How long is Google Cloud Professional Machine Learning Engineer Certification valid for?
2 years
Ml data and features¶
BigQuery (ml data and features)
Petabyte-scale analytics, structured data
Cloud Storage (ml data and features)
Object storage (CSV, JSON, Parquet, Avro, TFRecord)
Pub/Sub (ml data and features)
Real-time streaming data ingestion
Cloud SQL/Spanner (ml data and features)
Relational databases
Bigtable (ml data and features)
NoSQL for time-series and high-throughput workloads
Firestore (ml data and features)
Document database
External APIs (ml data and features)
RESTful services, partner data
Featurestore (ml data and features)
Container for all features
Entity Type (ml data and features)
Represents an object (e.g., user, product)
Feature (ml data and features)
Individual feature within an entity
Online Serving (ml data and features)
Low-latency feature retrieval for predictions
Offline Serving (ml data and features)
Batch feature retrieval for training
Point-in-time lookup (ml data and features)
Historical feature values at specific timestamps
Ml development and training¶
Grid Search (ml development and training)
Small parameter space, exhaustive search needed
Random Search (ml development and training)
Moderate parameter space, exploration
Bayesian (Vizier) (ml development and training)
Large parameter space, expensive training, best performance
BigQuery ML (ml development and training)
SQL users, data already in BigQuery, simple models
AutoML (ml development and training)
Quick POC, no ML expertise, standard problems
Agent Platform Managed Training (ml development and training)
Custom code, specific requirements, research
Pre-trained APIs (ml development and training)
No training data, standard tasks, immediate results
Answer (ml development and training)
AutoML Vision (8000 milli node hours budget)