Manav Goyal

Introduction - Artificial Intelligence

Classification
- Artificial Intelligence
  - Machine Learning => Supervised, Unsupervised, Reinforcement => Classification, Regression
    - Deep Learning => Artificial Neural Network => CNN, RNN (LSTM), Transformers, AutoEncorders
- Computer Vision => Image Processing
- Data Science => Big Data, IOT
- Human Computer Interaction => Augmented Reality, Virtual Reality, Telepresence, Robotics
Use Cases
- Face Recognition, Real-time Translation, Forecasting, DeepFake
- NLP
John McCarthy - 1956
- Machine to mimic human behavior
Machines
- Reasoning
  - Coming to logical conclusions
- Learning
  - Improving from experiences
- Problem Solving
  - Searching all possible outcomes to solve a problem
- Perception
  - Taking decisions bt self
State Space Search
- Number of states in which problem can go
- Representation
  - Precise
  - Analyze
- S: {S, A, Action (S), Result (S, a), Cost (S, a)}
  - S = Total number of states
    - Start, Goal and Intermediate States
  - A = Set of possible actions
  - Action (S) = Action done to reach new state (Searching for Goal state)
  - Result (S, a) = New state obtained after performing an action
  - Cost (S, a) = Cost of performing an action
Types of Search
- Uninformed (Blind)
  - Searching without information => Brute force method
  - Time consuming => Exponential Growth => Non Polynomial (NP) problem
    - More Costly
  - Time Complexity = O(Branching Factor^Depth)
  - Always gives Optimal solution
- Informed
  - Searching with information (Heuristic)
  - Quick Solution
  - Less Complexity => Both Time & Space
Knowledge Representation
- If not applied, can lead to Errors
  - Syntax
    - Grammatical error
  - Semantic
    - Logical error
- Theory
  - Logic
    - Propositional Logic
      - Syntax
        
        Atomic
        
        Only one sentence
        
        Complex/Composite
      - Semantic
    - Predicate Logic
  - Rules
    - if then
  - Semantic Net (Meaning Graph)
    - GoogleGraph
  - Frames
    - Slots (Objects)
    - Filters (Attributes)
  - Script
- Reasoning
Agents/Intelligent Agents
- Agent > Percept > Decision > Actions
Fuzzy Logic
- Basic
  - Lotfi Zadeh is the father of fuzzy logic
  - It is a superset of Boolean logic
  - Represent uncertainty (Continuous, not discrete)
  - Represent with degree
  - Represent the belonging of a member of a crisp set to fuzzy set
  - Membership Function
    - Membership Value => Between 0 and 1
- Operations
- Fuzzy Sets
  - Crisp
  - α-cut
- Fuzzy Sets
Planning
- Overview
- Hierarchical
- Goal Stack
Multi-agent System
- Types
- Properties
Problems
- Traveling Salesman Problem (TSP)
- 8-Puzzle Problem
- N-Queen Problem