Artificial Intelligence

AI Tools & Frameworks; Search Algorithms in AI; Knowledge Representation

C-CAT

AI Tools & Frameworks

Machine Learning Frameworks

FrameworkDeveloperLanguageUse Case
TensorFlowGooglePython, JS, C++Production ML/DL
PyTorchFacebook/MetaPythonResearch, Dynamic graphs
Keras(on TensorFlow)PythonHigh-level deep learning
Scikit-learnOpen sourcePythonClassical ML algorithms
XGBoostCommunityPython, RGradient boosting
LightGBMMicrosoftPythonFast gradient boosting

NLP Tools

ToolDescription
NLTKNatural Language Toolkit — Python
spaCyIndustrial-strength NLP library
Hugging FacePre-trained transformer models hub
GensimTopic modeling, word embeddings

Computer Vision Tools

ToolDescription
OpenCVReal-time computer vision library
YOLOReal-time object detection
Detectron2Facebook's object detection framework
MediaPipeGoogle's ML solution for live/streaming media

Cloud AI Platforms

PlatformProviderServices
SageMakerAWSComplete ML workflow
Vertex AIGoogle CloudML/AI platform
Azure MLMicrosoftEnterprise ML
IBM WatsonIBMEnterprise AI and NLP
DatabricksDatabricksSpark-based ML

Search Algorithms in AI

AI uses search algorithms to find solutions to problems by exploring a state space.

28.1 Uninformed Search (Blind Search)

No domain knowledge — explores states systematically.

AlgorithmStrategyComplete?Optimal?TimeSpace
BFSExplore level by levelYesYes (unit cost)O(b^d)O(b^d)
DFSExplore deep firstNoNoO(b^m)O(bm)
IDDFSDFS with increasing depth limitYesYesO(b^d)O(bd)
UCSExplore cheapest path firstYesYesO(b^d)O(b^d)

b = branching factor, d = solution depth, m = maximum depth

28.2 Informed Search (Heuristic Search)

Uses a heuristic function h(n) estimating distance to the goal.

AlgorithmEvaluation FunctionOptimal?
Greedy Best-Firstf(n) = h(n)No
A Search*f(n) = g(n) + h(n)Yes (admissible h)

A Search:*

  • g(n) = actual cost from start to node n
  • h(n) = estimated cost from n to goal (heuristic)
  • f(n) = g(n) + h(n) = total estimated path cost
  • A* is optimal when h(n) is admissible (never overestimates) and consistent

Example heuristics:

  • Straight-line distance (Euclidean) for navigation problems

Manhattan distance for grid-based problems

Knowledge Representation

AI systems need to represent knowledge about the world to reason and make decisions.

Forms of Knowledge Representation

TypeDescriptionExample
Logical Rules (Propositional/Predicate)IF-THEN rulesIF temperature > 100 THEN fever
Semantic NetworksGraph of concepts and relationships"Dog" IS-A "Animal"; "Dog" HAS "Fur"
FramesObject-oriented knowledge structuresFrame STUDENT {name, age, subject, grade}
OntologiesFormal knowledge hierarchiesMedical ontology (diseases, symptoms)
Production RulesExpert system rulesMYCIN medical diagnosis rules
Bayesian NetworksProbabilistic graphical modelsP(Disease

Knowledge Base vs Database

FeatureDatabaseKnowledge Base
ContentRaw data (facts)Structured knowledge + inference rules
QuerySQLInference engine / reasoning
ReasoningNoYes
ExamplesMySQL, PostgreSQLProlog, OWL/RDF, Neo4J

Continue learning

Related notes

Put this topic into timed practice

Open mock tests when you want full-exam pacing, or keep drilling in practice mode.