1.1 Basic AI Concept and Terms 1.1

Basic AI Terms Definition

• AI: creating systems that perform task that require human intelligence...
• ML: subset of AI systems learn from data without explicit programming
• deep learning: type of ML, large scale neural networks, multiple layers model complex data patterns
• neural networks: computational models inspired by human brain
• computer vision: machines understand interpret visual information
• NLP: interaction of computer with human languages
• model: trained rep of patterns in data for predictions and decisions
• algorithm: set of rules or math procedures to process data train models
• training and inferencing: teach ML model by feading it data and using trained model to make predictions on new unseen data
• bias: systematic errors in AI models
• fairness: AI models unbiased
• fit: how well ML models learn patterns in data
• LLM: DL model trained on vast amounts of text data to understand and generate human-like language

Hierarchically; AI > ML > DL > Gen AI. And based on Computational Power; CPU < GPU < TPU

• Central PU, Graphics PU, Tensor PU

Various types of inferencing

• Batch(offline): prediction are made on large set of data all at once
  • processing large amounts oof data periodically
• Real-Time(online): predictions are made instantly as new data comes
  • fast response time (milliseconds)
• Stream: continous flow of incoming data is processed in real-time, iot devices and social media analytics
  • handling continuos data streams
• Edge: model inferencing performed on local devices(edge devices) instead of centralised cloud servers
  • low-latency on-device processing
• Hybrid: combination of batch and real-time inference
  • mix of real-time and batch processing

Describe the different types of data in AI models (for example, labeled and unlabeled, tabular, time-series, image, text, structured and unstructured).

• Labeled (Supervised Data): data from predefined labels...used in supervised learning
• Unlabeled (Unsupervised Data): data without predefined labels...so AI gat too find patterns or groupings on its own...used in unsupervised tasks like clustering and anomaly detection
• Tabular:
• Time-series: datapoints collected over time at regular intervals...used in forcasting, anomaly detection and trend analysis
• Image: visual data format, png, jpeg...
• Text: written language data in documents, chats...used in NLP apps
• Structured: higly organised, fixed schema data
• Unstructured: no predefined structure...free-form texts
• Others: Semi-supervised (mix of labeled and unlabeled) Semi-structured (json/xml), Audio, Video

Describe supervised learning, unsupervised learning, and reinforcement learning.

• Supervised learning: model trained on labeled data (inputs have corresponding correct output)...so the model learns patterns and relationships between inputs and outputs...
  • Types:
    • classification: predicts discrete category eg. 'spam' vs 'not spam'
    • regression predicts continuous values eg. house price prediction
• Unsupervised learning:: model is trained on unlabeled data...so it has to find patterns and relationships without predifined outputs
  • Types:
    • Clustering: grouped similar datapoints
    • Association: ID relationships bwn variables
    • Anomaly Detection: detects unusual patterns
• Reinforcement learning: model learns by interracting with the environment...receiving rewards or penalties based on its actions...aims to maximise cummulative rewards over time