AI General Terms Flashcards
Improve base knowledge
Neural Network
a computer system modeled on the human brain and nervous system.
What is Grounding in AI?
The process of connecting language models to factual, verifiable information and real-world data sources, ensuring outputs are based on accurate, current information rather than training data alone.
How does Google Cloud Platform implement Grounding?
GCP implements grounding through:
Vertex AI’s built-in data connectors
Integration with enterprise data sources
Real-time data validation
Ability to connect to external APIs and databases
What is Tuning in machine learning?
The process of adjusting model parameters during training to optimize performance. Includes both automated parameter adjustment during training and manual hyperparameter optimization.
How does Fine Tuning differ from regular tuning?
The process of taking a pre-trained model and further training it on a specific dataset for a particular task or domain. This allows the model to specialize while retaining its base knowledge.
What does Model Management encompass in GCP?
Comprehensive oversight of ML models including:
Version control
Deployment management
Resource allocation
Access control
Model lifecycle management
What aspects of ML models should be monitored?
Key monitoring aspects include:
Model performance metrics
Prediction quality
Resource utilization
Data drift
Model drift
Latency and throughput
Error rates
What is Prompt Management?
The systematic approach to:
Creating and organizing prompts
Versioning prompt templates
Testing prompt effectiveness
Measuring prompt performance
Standardizing prompt patterns across applications
What should be tracked in model Notes and Status?
Critical tracking elements:
Training history
Performance metrics
Deployment status
Known issues
Update history
Dependencies
Production readiness
What should be included in ML model change tracking?
Essential elements to track:
Code changes
Data updates
Parameter modifications
Performance impact
Environmental changes
Deployment status
Rollback points
What is the Softmax function?
A mathematical function that converts a vector of numbers into a probability distribution. Commonly used in neural networks’ output layer for multi-class classification, ensuring all probabilities sum to 1.
What is RAG?
Retrieval-Augmented Generation
A technique that enhances language model responses by:
Retrieving relevant information from external sources
Incorporating this information into the generation process
Providing current and accurate information
Maintaining source attribution
What are RNNs and their use cases?
Neural networks designed for sequential data processing:
Maintains internal memory state
Processes sequences one element at a time
Suitable for time series, text, and speech
Can handle variable-length inputs
What are CNNs and their primary applications?
Neural networks specialized for processing grid-like data:
Excellent for image processing
Feature detection through convolution operations
Spatial hierarchy learning
Reduced parameter count compared to fully connected networks
What is a Vector in ML context?
A mathematical representation of data points:
Ordered array of numbers
Represents features in multi-dimensional space
Used for embeddings in ML models
Enables similarity comparisons
Like words from a page represented in numbers
Like a set of numbers representing a picture
0’s adn 1’s
What is Few Shot Learning?
The ability of a model to learn from very few examples:
Requires minimal training examples
Leverages transfer learning
Uses meta-learning techniques
Particularly useful for rare cases or new categories
hat is Sentiment Analysis?
ML technique to determine emotional tone in text:
Classifies text as positive, negative, or neutral
Uses NLP techniques
Can detect emotional nuances
Common in customer feedback analysis
What is a Tensor?
Multi-dimensional array used in ML:
Generalizes vectors and matrices
Basic data structure in deep learning
Represents complex data relationships
Core component in frameworks like TensorFlow
What is Apache Beam?
Unified programming model for:
Batch and streaming data processing
Portable across execution engines
Pipeline-based processing
Supported natively in GCP Dataflow
What is Apache Airflow?
Platform to programmatically author, schedule, and monitor workflows:
Creates DAGs of tasks
Manages task dependencies
Handles retry logic
Monitors execution
What is a DAG in data processing?
A workflow representation where:
Tasks are nodes
Dependencies are directed edges
No cycles allowed
Defines processing order
What is Data Fabric?
Architecture that:
Integrates data sources
Provides unified data management
Enables consistent security
Supports data governance
What is a Data Lake?
Storage repository that:
Holds raw data in native format
Supports structured and unstructured data
Enables big data analytics
Scales horizontally
What is a Lake House architecture?
Hybrid architecture combining:
Data lake flexibility
Data warehouse performance
ACID transactions
Schema enforcement when needed
What are Hyperparameters?
Configuration settings used to control the learning process:
Set before training begins
Not learned from data
Examples: learning rate, batch size
Tuned through experimentation
