Genetic Algorithm (GA)

│
├── 1. Inspiration
│ └── Natural Evolution (Darwin's "Survival of the fittest")
│
├── 2. Basic Terminologies
│ ├── Chromosome → Solution (string)
│ ├── Gene → Element of chromosome
│ ├── Population → Set of chromosomes
│ ├── Fitness Function → Evaluates how "good" a solution is
│
├── 3. GA Cycle (Main Steps)
│ ├── a) Initialization → Create initial population randomly
│ ├── b) Selection → Choose best individuals (parents)
│ ├── c) Crossover → Mix genes of parents to create children
│ ├── d) Mutation → Random tweak to maintain diversity
│ └── e) Replacement → Replace old generation with new
│
├── 4. Encoding Strategies
│ ├── Binary Encoding → 1100101
│ ├── Permutation Encoding → Order-based (TSP problems)
│ ├── Value Encoding → Real numbers (e.g., [3.2, 7.1])
│ └── Tree Encoding → Used in Genetic Programming
│
├── 5. Selection Methods
│ ├── Roulette Wheel Selection
│ ├── Tournament Selection
│ ├── Rank Selection
│ └── Elitism
│
├── 6. Crossover Techniques
│ ├── Single-Point Crossover
│ ├── Two-Point Crossover
│ └── Uniform Crossover
│
├── 7. Mutation Techniques
│ ├── Bit Flip (Binary)
│ ├── Swap Mutation (Permutation)
│ └── Gaussian Mutation (Real-Value)
│
├── 8. Termination Conditions
│ ├── Reached max generations
│ ├── Fitness threshold met
│ └── No improvement for N iterations
│
└── 9. Applications
├── Optimization Problems
├── Scheduling & Routing
├── Neural Network Weight Training
└── Robotics & Game AI
Artificial Neural Networks (ANN)
│
├── 1. Learning Types
│ ├── Supervised Learning
│ │ └── Input-output pairs provided
│ ├── Unsupervised Learning
│ │ └── No target output, clustering & pattern discovery
│ └── Reinforcement Learning
│ └── Feedback based on actions (reward/punishment)
│
├── 2. Basic Concepts
│ ├── Neuron (Processing Element)
│ │ └── Takes weighted input, applies activation
│ ├── Weights → Strength of connection
│ ├── Bias → Threshold adjustment
│ ├── Activation Function
│ │ ├── Step Function
│ │ ├── Sigmoid Function
│ │ ├── ReLU (Rectified Linear Unit)
│ │ └── Tanh
│
├── 3. Single-Layer Perceptron
│ ├── One input layer, one output
│ ├── Can solve only *linearly separable* problems
│ └── Uses Perceptron Learning Rule
│
├── 4. Multi-Layer Perceptron (MLP)
│ ├── Input Layer → Hidden Layer(s) → Output Layer
│ ├── Can solve *non-linearly separable* problems
│ └── Learns using *Backpropagation Algorithm*
│
├── 5. Self Organizing Maps (SOM)
│ ├── Unsupervised learning
│ ├── Maps high-dimensional data to lower dimensions
│ └── Preserves topological properties
│
├── 6. Hopfield Network
│ ├── Recurrent Neural Network (RNN)
│ ├── Symmetric weights (no self-connection)
│ ├── Used for associative memory
│ └── Converges to stable state
│
└── 7. Applications
├── Pattern Recognition
├── Image & Speech Processing
├── Medical Diagnosis
├── Stock Market Prediction
└── Robotics & Control Systems

Fuzzy Sets
│
├── 1. Basic Concepts

│ ├── Membership Function μ(x) ∈ [0, 1]

│ ├── Fuzziness → Partial truth (not just 0 or 1)

│ ├── Core → x where μ(x) = 1

│ ├── Support → x where μ(x) > 0
│ ├── Height → max(μ(x))
│ ├── Normal Set → At least one μ(x) = 1
│ └── Subnormal Set → All μ(x) < 1
│
├── 2. Membership Function Types
│ ├── Triangular (△)
│ ├── Trapezoidal (▭)
│ ├── Gaussian (bell-shaped)
│ └── Sigmoid (S-shaped)
│

│ ├── Union (A ∪ B) → max(μA, μB)

├── 3. Set Operations

│ ├── Intersection (A ∩ B) → min(μA, μB)

│ └── Complement (¬A) → 1 – μA(x)
│
├── 4. Fuzzification
│ ├── Crisp input → Fuzzy value
│ └── Example: Temp 70°C → “Hot” with μ=0.9
│
├── 5. Rule Base
│ ├── IF–THEN Format
│ └── Example: IF speed is high THEN brake is strong
│
├── 6. Inference Engines
│ ├── Mamdani → Output: fuzzy set (needs defuzzification)
│ └── Sugeno → Output: crisp value or function (no defuzz needed)
│
├── 7. Defuzzification Methods
│ ├── Centroid (Center of Area) → Most common
│ ├── Bisector
│ ├── Mean of Maximum (MoM)
│ ├── Smallest of Maximum (SoM)
│ └── Largest of Maximum (LoM)
│
└── 8. Applications
├── Washing Machines
├── Air Conditioners
├── Smart Traffic Lights
├── Decision Support Systems
└── Medical Diagnosis

Multi-Agent Systems (MAS)

│
├── 1. Agents and Objects
│ ├── Agent = Autonomous entity
│ ├── Object = Passive data structure (no control)
│ └── Difference → Agents can decide, act, learn
│
├── 2. Agents vs Expert Systems
│ ├── Expert System → Rule-based, single agent, static
│ └── MAS → Multiple agents, interaction, dynamic behavior
│
├── 3. Structure of MAS
│ ├── Multiple Agents
│ ├── Shared Environment
│ ├── Communication Protocol
│ ├── Coordination Mechanism
│ └── Goal-Oriented Behavior
│
├── 4. Semantic Web
│ ├── Adds meaning to data on web
│ ├── Based on Ontologies
│ └── Supports Agent Communication
│
├── 5. Agent Communication
│ ├── ACL (Agent Communication Language)
│ ├── KQML (Knowledge Query Manipulation Language)
│ ├── Speech Acts → Request, Inform, Query, etc.
│ └── Performatives = Message types
│
├── 6. Knowledge Sharing via Ontologies
│ ├── Ontology = Shared vocabulary of concepts
│ ├── Allows interoperability between agents
│ └── Example Tool: Protégé (ontology editor)
│
├── 7. Agent Development Tools
│ ├── JADE (Java Agent DEvelopment)
│ ├── FIPA compliant
│ └── Others: ZEUS, AgentBuilder
│
└── 8. Applications of MAS
├── Distributed Problem Solving
├── E-Commerce (Bidding, Negotiation)
├── Robotics (Swarm Intelligence)
├── Smart Grid & Energy Systems
└── Traffic Control & Simulation

Natural Language Processing (NLP)

│
├── 1. Basics of NLP
│ ├── Study of human (natural) language by computers
│ ├── Input: Speech/Text → Processing → Output: Meaning/Action
│ └── Examples: ChatGPT, Google Translate, Siri

├── 2. Components of NLP

│
│ ├── a) Lexical Analysis
│ │ └── Breaking text into tokens (words, punctuation)
│
│ ├── b) Syntax Analysis (Parsing)
│ │ ├── Grammar-based analysis
│ │ ├── Checks sentence structure
│ │ └── Uses Parse Trees, CFGs (Context-Free Grammar)
│
│ ├── c) Semantic Analysis
│ │ ├── Assigns meaning to words/sentences
│ │ ├── Maps words to concepts
│ │ └── Resolves ambiguity using context
│
│ ├── d) Pragmatics
│ │ ├── Understanding context and speaker's intent
│ │ └── Example: “Can you pass the salt?” = a request, not question
│
│ └── e) Discourse Analysis
│ ├── Understands flow across multiple sentences
│ └── Tracks topics, coherence, reference (he, she, it)

├── 3. NLP Tasks

│ ├── Tokenization
│ ├── POS Tagging (Noun, Verb, etc.)
│ ├── Named Entity Recognition (NER)
│ ├── Chunking
│ ├── Parsing
│ ├── Lemmatization / Stemming
│ ├── Sentiment Analysis
│ └── Machine Translation

├── 4. Techniques in NLP

│ ├── Rule-Based → Grammar + Dictionaries
│ ├── Statistical → Based on probabilities (n-grams, HMMs)
│ └── Machine Learning → Deep learning, Transformers, BERT

├── 5. Grammar in NLP

│ ├── Phrase Structure Grammar (PSG)
│ ├── Dependency Grammar
│ └── Chomsky Hierarchy relevance: CFGs used for parsing

├── 6. Ambiguity in NLP

│ ├── Lexical Ambiguity → One word, many meanings (e.g., "bank")
│ ├── Syntactic Ambiguity → Sentence structure confusion
│ └── Semantic Ambiguity → Conflicting meaning based on context
├── 7. Applications of NLP
│ ├── Chatbots, Virtual Assistants
│ ├── Machine Translation (Google Translate)
│ ├── Text Summarization
│ ├── Speech Recognition
│ └── Sentiment Analysis (e.g., Twitter, Reviews)

└── 8. NLP Libraries & Tools

├── NLTK (Python)
├── spaCy
├── GPT / BERT / Transformers
├── Stanford NLP
└── OpenNLP

AI Planning
│
├── 1. What is Planning?
│ ├── Process of deciding a sequence of actions
│ ├── Goal: Reach desired goal state from initial state
│ └── Used in robotics, games, logistics, etc.
│
├── 2. Planning Problem
│ ├── Initial State
│ ├── Goal State
│ ├── Actions (Operators)
│ ├── State Space
│ └── Plan = Sequence of actions from start to goal
│
├── 3. Types of Planning
│
│ ├── a) Classical Planning
│ │ ├── Fully observable environment
│ │ ├── Deterministic actions
│ │ └── Example: STRIPS, Situation Calculus
│
│ ├── b) Conditional Planning
│ │ ├── Uncertainty in action outcomes
│ │ └── Includes IF-THEN-ELSE conditions
│
│ ├── c) Contingency Planning
│ │ ├── Plans for alternative possibilities
│ │ └── Used in real-world planning (e.g., GPS rerouting)
│
│ ├── d) Hierarchical Task Network (HTN) Planning
│ │ ├── Top-down decomposition of tasks
│ │ └── Used in practical applications (logistics, robotics)
│
│ └── e) Probabilistic Planning
│ ├── Outcomes have probabilities
│ ├── Uses MDP (Markov Decision Process)
│ └── Used in reinforcement learning
│
├── 4. Planning Techniques / Representations
│ ├── STRIPS (Stanford Research Institute Problem Solver)
│ ├── State-Space Search (Forward & Backward)
│ ├── Planning Graph (used in Graphplan)
│ ├── Situation Calculus (Logic-based planning)
│ └── Partial Order Planning (POP)
│
├── 5. Plan Execution & Monitoring
│ ├── Match actual state with expected state
│ ├── Replan if deviations occur
│ └── Used in real-time AI agents
│
└── 6. Applications of Planning
├── Robot Navigation
├── Automated Logistics
├── Game AI (Chess, Strategy Games)
 ├── Workflow Management
└── Smart Assistants (Goal-directed tasks)

Knowledge Representation (KR)

│
├── 1. What is KR?
│ ├── Process of encoding real-world knowledge into a format that machines can understand
│ ├── Converts human knowledge → machine-usable form
│ └── Foundation of intelligent reasoning
│
├── 2. Desirable Properties of KR
│ ├── Representational Adequacy (Expressive power)
│ ├── Inferential Adequacy (Support reasoning)
│ ├── Inferential Efficiency (Quick reasoning)
│ └── Acquisitional Efficiency (Easy to acquire knowledge)
│
├── 3. Types of Knowledge
│ ├── Declarative → Facts (e.g., Delhi is capital of India)
│ ├── Procedural → How-to (e.g., how to drive a car)
│ ├── Meta-Knowledge → Knowledge about knowledge
│ ├── Heuristic Knowledge → Experience-based rules
│ └── Structural Knowledge → Relationships (e.g., hierarchies)
│
├── 4. Approaches/Techniques of KR
│
│ ├── a) Logic-Based Representation
│ │ ├── Propositional Logic
│ │ ├── Predicate Logic (First Order Logic)
│ │ └── Enables inference using rules
│
│ ├── b) Semantic Network
│ │ ├── Graph-based → Nodes (concepts) + Edges (relations)
│ │ ├── Example: “Bird → isA → Animal”
│ │ └── Easy to visualize relationships
│
│ ├── c) Frame-Based Representation
│ │ ├── Data structure with slots and values
│ │ ├── Like object-oriented model
│ │ └── Example: Frame(Car): {Wheels=4, Engine=Yes}
│
│ ├── d) Production Rules (IF-THEN)
│ │ ├── Rule-based reasoning
│ │ └── Example: IF it rains THEN carry umbrella
│
│ └── e) Ontology-Based KR
│ ├── Defines domain concepts and relationships
│ ├── Machine-interpretable
│ └── Used in Semantic Web
│
├── 5. Reasoning Types in KR
│ ├── Deductive Reasoning → General → Specific (logic)
│ ├── Inductive Reasoning → Specific → General (learning)
│ ├── Abductive Reasoning → Inference to best explanation
│ └── Default Reasoning → Assume unless told otherwise
│
└── 6. Applications
├── Expert Systems
├── Medical Diagnosis (e.g., MYCIN)
├── Natural Language Understanding
├── Robotics
 └── Chatbots & Virtual Assistants

Approaches to Artificial Intelligence (AI)

│
├── 1. Logical Approach
│ ├── Based on formal logic
│ ├── Uses rules and facts
│ └── Example: Propositional and Predicate Logic
├── 2. Knowledge-Based Approach
│ ├── Stores expert knowledge
│ ├── Uses inference engine
│ └── Example: Expert Systems (MYCIN)

├── 3. Search-Based Approach

│ ├── Explores state space for solution
│ ├── Uses algorithms like BFS, DFS, A*
│ └── Example: Path finding in games

├── 4. Planning Approach

│ ├── Decides sequence of actions
│ ├── Based on initial and goal states
│ └── Example: STRIPS, Situation Calculus

├── 5. Machine Learning Approach

│ ├── Learns from data
│ ├── Includes supervised, unsupervised, reinforcement learning
│ └── Example: Neural Networks, SVM, Decision Trees

├── 6. Neural Network Approach

│ ├── Mimics human brain
│ ├── Uses layers of artificial neurons
│ └── Example: Perceptron, MLP

├── 7. Evolutionary Approach

│ ├── Inspired by natural evolution
│ ├── Uses Genetic Algorithms (GA)
│ └── Applies crossover, mutation, selection

├── 8. Reactive/Behavior-Based Approach

│ ├── Focuses on real-time behavior
│ ├── Doesn't use internal models
│ └── Example: Subsumption architecture (Rodney Brooks)

└── 9. Hybrid Approach

├── Combines multiple approaches
├── For flexibility and accuracy
└── Example: Neuro-Fuzzy Systems, Deep Reinforcement Learning