Artificial Intelligence
 

Syllabus

 
 
 
UNIT 1
Course outcomes and Syllabus

What is Intelligence ?

  • Capacity for logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning and problem solving. More generally it can be described as the ability to perceive information, and to retain it as knowledge to be applied towards behaviors within an environment.
  • Intelligence is a property of mind that encompasses many related mental abilities, such as the capabilities to
    • reason
    • plan
    • solve problems
    • think abstractly
    • comprehend ideas and
    • language and learn

What is Artificial Intelligence ?

  • Artificial intelligence refers to the simulation of human intelligence in machines. The goals of artificial intelligence include learning, reasoning, and perception.
Formal definition of AI
  • AI is a branch of computer science which is concerned with the study and creation of computer systems that exhibit some form of intelligence OR those characteristics which we associate with intelligence in human behavior.
  • AI is a broad area consisting of different fields, from machine vision, expert systems to the creation of machines that can "think". In order to classify machines as "thinking", it is necessary to define intelligence.
 
Categories and Foundation of AI

Categories of AI Systems

Systems that think like Humans
  • Black box. Thought process not clear
  • Area of Cognitive Science
    • Stimuli —> mental representation
  • Neural Network
Systems that act like Humans
  • Turing test to see if AI behaves like human. 1950.
  • Overall behaviour of system should be human like
  • Could be achieved by observation
  • Chatbots etc.
Systems that think rationally/logically
  • Logic or laws of thoughts
  • Logic formulas, prepositional logic are used for synthesizing outcomes
  • Not all intelligent behaviour are mediated by logical deliberation
Systems that act rationally/logically
  • agent
  • Rational behaviour is doing right thing
  • Even if method is illogical, observed behaviour must be rational .
 

Foundations of AI

Philosophy
  • Where does knowledge come from?
  • How does knowledge lead to action?
Mathematics
  • Formal logic methods
    • Boolean logic
    • Fuzzy logic
  • Uncertainity
    • Basis for modern approach to handle uncertainity can be handled by
      • Probability Theory
      • Modal and Temporal Logics
Economics
  • Decisions to maximize payoff
  • Decision to take if payoff far away
  • Optimizing certain values
Neuroscience
  • How brain works?
    • 1824 studies on abnormal people
    • Recent studies accurate sensors to measure brain waves
    • Parallel computing, remapping, interconections, ANN, NNN
Control Theory
  • Machine modify behaviour in response to action
    • Water flow regulator
    • Steam engine generator
  • Theory of stable feedback systems
    • Transition from initial state to goal state with minimum energy
    • Control theory could only describe linear systems
Linguistics
  • How does language relate to thought
  • Speech depicts human intelligence
  • Children make sentences they never heard before
  • Language and thought go hand in hand
Computer Engineering
  • Building efficient computers
  • Application specific
  • Quantum computing
Psychology
  • How do humans think and act
History, Sub-areas and applications of AI

History

  • 1943 - Artificial Neurons - Warren & Walter
  • 1949 - Hebbian Learning - Donald Hebb
  • 1950 - Turing Test - Alan Turing
  • 1955 - Logic Theorist - Allen & Helbert
  • 1956 - John McCarthy - AI made an academic field
    • High level computer languages COBOL FORTRAN LISP created
  • 1966 - ELIZA (first chatbot) - Joseph
  • 1972 - Intelligent Humanoid Robot WABOT-1 - made in Japan
  • 1980 - Expert Systems - Emulate decision making ability of human
    • First National conference of AI at stanford
  • 1997 - IBM Deep Blue beats world chess champion
  • 2002 - Roomba - AI enters home
  • 2006 - FAANG - AI enters business
  • 2011 - IBM Watson wins quiz show - NLP
  • 2012 - Google Now - predict according to user info
  • 2014 - Eugene Goostman chatbot - came very close to pass turing test

Sub-areas of AI

  • Neural Networks
    • Inspired by human brain
  • Vision
  • Machine Learning
  • Speech Recognition
  • Robotics

Applications

  • Business - Financial Strategies
  • Engineering - Design, offer strategy
  • Manufacturing - Assembly, inspection
  • Mining
  • Medical
  • Education
  • Housing
  • Farming
Stucture of Agents and Problem Solving

Agent

  • An agent perceives its environment via sensores and acts upon that environment through its actuators
  • AI can be defined as a study of rational agents

Rational Agent

  • An agent could be anything which makes decisions like a person, firm or software. It carries out an action with the best outcome after considering past and current percepts(agent’s perceptual inputs at a given instance)

An AI System is composed of two main components :

  • Agent
  • Environment

Structure of Agents

Agent = Architecture + Agent Program
Architecture : the machinery that the agent executes on. It is a device with sensors and actuators, for example : a robotic car
Agent Program : Agent program is an implementation of an agent function. An agent function is a map from the percept sequence(history of all that an agent has perceived till date) to an action.
Examples
  • A software agent has Keystrokes, file contents, received network packages which act as sensors and displays on the screen, files, sent network packets acting as actuators.
  • A Human agent has eyes, ears, and other organs which act as sensors and hands, legs, mouth, and other body parts acting as actuators.
  • A Robotic agent has Cameras and infrared range finders which act as sensors and various motors acting as actuators.

Types of Environment

  • Fully Observable : Agent's sensors give it access to complete state of environment
  • Partially Observable : Either noisy/inaccurate sensors or parts of the state are missing from sensor data.

Defining Task Environment

PEAS
Ex : Self driving car
  • Performance Measure : Safe, Fast, Legal
  • Environment : Roads, other traffic
  • Actuators : steering wheel, brake, signal
  • Sensors : Cameras, LIDAR, GPS, odometer etc.

Types of Agents

Simple Reflex Agents
  • Act only on current percept. Ignore past percept.
Agent functions is based action-condition rule
A condition-action rule is a rule that maps a state i.e, condition to an action. If the condition is true, then the action is taken, else not.
  • Works only in fully observable environment
  • Cons
      1. Limited Intelligence
      1. No knowledge of non-perceptual parts of state
      1. Usually too big to generate and score
      1. If any update in environment, rules have to be changed
notion image
Model-based Agents
  • Works by finding a rule whose condition matches the current situation
  • Can work in partially-observable environment by use of model about the world
  • Agent keeps track of internal state adjusted by each percept and Depends on percept history
  • Two very important parts
    • Model : knowledge about "how things happen in the world"
    • Internal State: It is a representation of the current state based on percept history
  • These agents have the model, "which is knowledge of the world" and based on the model they perform actions.
  • Updating agent state requires information about
    • How the world evolves
    • How the agent actions effect the world
notion image
Goal-based Agents
  • Expand on the capabilities of a Model-based agent by having the goal information
  • The choose actions so that they can achieve the goal
  • Agents may have to consider a long sequence of possible actions before deciding whether the goal is achieved or not. Such considerations of different scenario are called searching and planning, which makes an agent proactive.
notion image
Utility-based Agents
  • Expand on the capabilities of a Goal-based agent by providing an extra component of Utility Measurement which provides measure of success for every state
  • They act based not only on goals, but also strive to obtain the best path of reaching the goal
  • Useful when there are multiple alternatives of reaching goal
  • Maps each state to a real number to check efficiency of action in achieving goal or 'success' rate
notion image
Learning Agent
  • Agent which can learn from it's past experiences
  • Starts with basic knowledge and then able to act and adapt automatically by learning
  • Four main components
      1. Learning Element : Makes improvements by learning from environment
      1. Critic : Learning Element takes feedback from critic to know how well it's doing compared to a standard performance
      1. Performance Element : Responsible for selecting external action
      1. Problem Generator : Responsible for suggesting actions that will lead to new/informative experiences
notion image

Formulating or Defining a Problem

  • To solve a problem defining it clearly is very important
A problem can be defined formally by five components
PITAG
  1. Initial State
    1. The state where agents start to perform the search to reach the goal state
  1. Actions
    1. Set of applicable actions to perform in state S
  1. Transition Model
    1. What each actions does. Transition each actions does.
  1. Goal Test
    1. This determines whether a given state is Goal Test or not.
  1. Path Cost
    1. Assigns numeric cost to each path
Problem Types
  • Single - State Problem
      1. Observable
      1. Deterministic
      1. static
      1. discrete
  • Multi - State Problem
      1. Partially observable
      1. Deterministic
      1. static
      1. discrete
  • Contingency Problem
      1. Partially Observable
      1. Non - Deterministic

State Space

  • No.of states the problem can go
    • Why is it needed ?
    • Analyzing problem
    • Representing the problem precisely
  • Together the initial state, actions, transition model implicitly define the state space of the problem. Set of all states reachable from initial state by any sequence of actions.
  • When graphically represented, state space forms directed graph in which nodes are states and edges are actions
  • A path is thus a sequence of state and sequence of actions taken to reach those states
  • Process of removing detail from a representation is called Abstraction
 

What is Heuristic ?

  • Technique designed to solve a problem quickly
  • Used to convert NP problem to P. [ NP = Non-polynomial ]
  • Gives good solution but does not guarentee optimal

Problems on State Space Representation

The 8 - Puzzle
  • We are given Initial State and Goal State
  • States : Individual Tiles
  • Actions : Move Up, Down, Left or Right
  • Goal Test := Compare with goal state
  • Path Cost : 1 per move
  • 8 Puzzle has 9!/2 = 181440 states
notion image
There are two ways of solving this problem :
  • Uninformed Search ( Breadth First Search or DFS )
    • Just do a normal BFS where ALL possible states are searched and most optimal one is returned
    • Time Complexity : O ( BranchingFactor ^ DepthFactor ) for this problem its [ 3 ^ 20 ]
  • Informed / Heuristic Search ( Best First Search , A* Algo )
      1. Find Heuristic Value ( R )
        1. This value will consist of misplaced tiles when current state (initial state here) is compared to goal state [which is 3 in our case here]
      1. Start Exploring all legal moves (up,down,left,right while in boudary)
      1. Find Heuristic Value of all Legal Moves
      1. Continue exploring with minimum heuristic value since we want ideally 0 misplaced tiles
       
       
 
 
Search Algorithms and Strategies

Search Strategies

Adversial Search
 
 
Evaluation Function
An evaluation function
also known as
  • heuristic evaluation function
  • static evaluation function
is a function used by game-playing computer programs to estimate the value or goodness of a position (usually at a leaf or terminal node) in a game tree.
A tree of such evaluations is usually part of a minimax or related search paradigm which returns a particular node and its evaluation as a result of alternately selecting the most favorable move for the side on move at each ply of the game tree.
The value is a quantized scalar,
The value is presumed to represent the relative probability of winning if the game tree were expanded from that node to the end of the game
Write the python code instead of illegible pseudo code
Mini - Max Algo Code [Python]
# A simple Python3 program to find
# maximum score that
# maximizing player can get
import math

def minimax (
							curDepth, 
							nodeIndex,
							maxTurn, 
							scores,
							targetDepth
						):

	# base case : targetDepth reached
	if (curDepth == targetDepth):
		return scores[nodeIndex]
	
	if (maxTurn):
		return max(
								minimax(
													curDepth + 1, 
													nodeIndex * 2,
													False, 
													scores, 
													targetDepth
												),
								minimax(
													curDepth + 1, 
													nodeIndex * 2 + 1,
													False, 
													scores, 
													targetDepth
												)
							)
	
	else:
		return min(minimax(curDepth + 1, nodeIndex * 2,
					True, scores, targetDepth),
				minimax(curDepth + 1, nodeIndex * 2 + 1,
					True, scores, targetDepth))
	
# Driver code
scores = [3, 5, 2, 9, 12, 5, 23, 23]

treeDepth = math.log(len(scores), 2)

print("The optimal value is : ", end = "")
print(minimax(0, 0, True, scores, treeDepth))

# This code is contributed
# by rootshadow
[OPTIMIZED] Mini - Max Algo with alpha-beta pruning [Pseudo Code]
The moment we encounter a Beta(Minimizer) which is less than Alpha (Maximizer), we automatically prune it because Apha won't chose anything less than or equal to it simply because it is a MAXIMIZER
function minimax(node, depth, isMaximizingPlayer, alpha, beta):

    if node is a leaf node :
        return value of the node
    
    if isMaximizingPlayer :
        bestVal = -INFINITY 
        for each child node :
            value = minimax(node, depth+1, false, alpha, beta)
            bestVal = max( bestVal, value) 
            alpha = max( alpha, bestVal)
            if beta <= alpha:
                break
        return bestVal

    else :
        bestVal = +INFINITY 
        for each child node :
            value = minimax(node, depth+1, true, alpha, beta)
            bestVal = min( bestVal, value) 
            beta = min( beta, bestVal)
            if beta <= alpha:
                break
        return bestVal
  • UNIT 2 skip ?
UNIT 3
Expert Systems and its Applications
Side Headings :
Introduction
  • Expert System is a computer program
  • Designed to provide Decision Making Ability (like a human expert)
  • Provides this ability by extracting knowledge from its knowledge base (using reasoning, inference rules, facts and heuristics (just like a human) on the User Query/input )
  • First system built in 1970
  • These systems are designed for specific domains such as medicine, law etc.
  • Performance of the system depends on quality/quantity of knowledge base
  • Suggestion of correct spelling in google is a well-known example of its use
notion image
 
Popular Examples
DENDRAL
  • Chemical Analysis Expert System
  • Used in organic chemistry to detect unknown organic molecules with the help of their mass spectra and knowledge base of chemistry.
MYCIN
  • Find bacteria causing infections like bacteraemia and meningitis
  • also used for the recommendation of antibiotics and the diagnosis of blood clotting diseases.
PXDES
  • determine the type and level of lung cancer.
  • Takes picture which identifies the type and degree of harm using knowledge base
CaDeT
  • Diagnostic support system that can detect cancer at early stages
 
Characteristics
1. High Performance
The expert system provides high performance for solving any type of complex problem of a specific domain with high efficiency and accuracy.
2. Understandable
It responds in a way that can be easily understandable by the user. It can take input in human language and provides the output in the same way.
3. Reliable
It is much reliable for generating an efficient and accurate output.
4. Highly Responsive
ES provides the result for any complex query within a very short period of time.
 
Components
notion image
User Interface
Interface helps a non-expert user to communicate with the expert system to find a solution.

Responsibilities

  • Accepting query in readable format
  • Sanitization of Query
  • Displays output to user
Inference Engine
  • Brain of the Expert System as it is the main processing unit of the system
  • ⭐ It applies Inference rules to the knowledge base to extract an answer or conclusion
Types of Inference Engines
Deterministic Inference engine
  • Conclusions drawn are assumed to be true.
  • Based on facts and rules.
Probabilistic Inference engine
  • Conclusions contain uncertainity
  • Conclusions are based on probability
Modes of Deriving Solution
Forward Chaining
  • Starts with known facts and rules, applies inference rules to get conclusion and spits output.
notion image
Backward Chaining
  • Backward reasoning method
  • Starts from the goal and works backward to prove the known facts.
notion image
Knowledge Base
  • Type of storage that stores knowledge acquired from the different experts of the particular domain.
  • It can also be viewed as collection of Objects and their Attributes
    • Ex : Lion (object) is a Mammal (Attribute)
Components of Knowledge Base
  • Factual Knowledge
    • Based on facts and accepted by knowledge engineers
  • Heuristic Knowledge
    • Based on practice, the ability to guess, evaluation, and experiences.
Resources
 
Phases in Building
Participants in Development
Domain Experts
The success of an ES much depends on the knowledge provided by human experts. These experts are those persons who are specialized in that specific domain.
Knowledge Engineers
Knowledge engineer is the person who gathers the knowledge from the domain experts and then encodes that knowledge to the system according to the formalism.
End - Users
This is a particular person or a group of people who may not be experts, and working on the expert system needs the solution or advice for his queries, which are complex.
Short - Explanation
There are five phases of processing of Expert system, such as-
Determine Problem Domain
  • Expert system must be bearing the problems which are solved by it.
  • Assign the domain expert for that ES.
  • Then, to build expert system that is cost- effective.
Design System
  • To build the high degree of integration this is able for another system and database.
  • Realize how the concepts can represent the domain knowledge best.
Build Prototype
  • To complete this phase, need knowledge expert who is able to do these tasks, such as-
  • Need domain knowledge from expert side.
  • And, then to represent this system into If-THEN-ELSE rules.
Testing the Prototype
  • End users allow testing the build prototypes of Expert system.
Development of Production Grade System
  • After testing, ensure that all interactions of expert system along with their elements of its environment, end users, all databases, and another system as well.
  • To keep all documents of the expert system assignment
Maintain
  • To update of ES continuously up to date.
 
Resources
Long - Explanation
 
 
notion image
 
Identification [ Understand Problem ]
  • Precisely describe :
    • Knowedge Domain
    • Exact nature of the problem which system is intended to solve
    • State the precise goals that indicate exactly how system will contribute to solution

Iterative Procedure of improving Problem description in Engineer's head

  1. Knowledge Engineer gains familiarity with that domain
  1. Domain Expert describes several common problem states
  1. Knowledge Engineer tries to extract fundamental concepts for those problem states
  1. Domain Expert suggests changes ( Break; if description perfect. Fire ; if infinite loop )
  1. Back to Square 1
 
notion image
 
Questions to ask
  • Is one Domain Expert enough ?
  • Should we consult more resources (books, research papers, scientists etc.) ?
  • Is a single engineer enough ?
  • Any deadlines ?
 
Conceptualization [ Conceptualize Problem ]

Purpose :

  • Analyse problem further to ensure specifics and generalities understood

Iterative procedure of conceptualizing problem in Engineer's head

  • Knowledge engineer draws out relation between objects and processes in problem domain
  • Divides problem into sub-problems
  • Create proper relationships between those sub-problems
  • Communicate with Domain Expert and make sure relations, objects and processes are as intended
  • Loop again if not ended
 
 
notion image
 
Note : If any weaknesses of previous stages are discovered, we go back to that stage add it and start again from that stage. Classic iterative procedure.
Formalization [ Problem is Connected to Proposed Solution]
  • Most interactive and time consuming stage
  • Not unusual for this stage to last for several years
  • Expert system is supplied by analyzing the relationships depicted in the last stage (conceptualization stage).
  • The knowledge engineer begins to select the techniques which are appropriate for developing this particular expert system.
Knowledge Engineer should be familier with
  • The various techniques of knowledge representation and intelligent search techniques used in expert systems.
  • The expert system tools which can greatly expedite the development process.
  • Other expert systems which may solve similar problems and thus may be adaptable to problem at hand.
 
Example : In rule based system
  • Engineer Develops set of rules to represent knowledge communicated by Expert
  • Many domain experts know "what" they do but not "why" they do it. This is a challenge for the engineer that he/she has to overcome
  • Domain Experts reviews rules
    • IF okay
      • Break ;
    • ELSE
      • Iterate again with suggestions
 
 
Implementation [ formalised concepts are programmed into the computer to build a prototype ]
  • Formalised concepts are programmed into the computer which has been chosen for system development using the predetermined techniques and tools to implement a ‘first-pass’ (prototype) of the expert system.
  • PROLOG was used as a programming language
  • The knowledge engineer may discover that the chosen techniques simply cannot be implemented.
  • It may not be possible, for example, to integrate the knowledge representation techniques selected for different sub-problems.
  • At that point, the concepts may have to be re-formalised, or it even may be necessary to create new development tools to implement the system efficiently. Basically go back and change stuff to make this stage work.
 
notion image
 
  • Once the prototype system has been refined sufficiently to allow it to be executed, the expert system is ready to be tested thoroughly to ensure that it expertise’s correctly.
 
Testing [ Identify weakness in Structure and Implementation to make appropriate corrections ]
 
  • Once the system has proven to be capable of correctly solving straight-forward problems, the domain expert suggests complex problems which typically would require a great deal of human expertise. These more demanding tests should uncover more serious flaws and provide ample opportunity to ‘fine tune’ the system even further.
  • Ultimately, an expert system is judged to be entirely successful only when it operates at the level of a human expert. The testing process is not complete until it indicates that the solutions suggested by the expert system are consistently as valid as those provided by a human domain expert.
 
 
 
Resources
 
Reading directly from PDF
 
Architecture
 
notion image
 
Knowledge Base
  • Warehouse of special heuristics or rules used directly by knowledge, facts (productions)
  • Has knowledge that is needed for understanding, formulating, & problem solving.
Working Memory
  • describes current running problem and record intermediate output.
  • Records Intermediate Hypothesis & Decisions in following pattern:
      1. Plan,
      1. Agenda
      1. Solution
Inference Engine
  • Heart of expert system
  • Helps to manage entire structure of expert system
  • Uses different methodology for reasoning.
Knowledge Acquisition System
  • Knowledge acquisition is the process used to define the rules required for a knowledge-based system.
  • It is the process of extracting, organizing, and structuring the domain knowledge, specifying the rules to acquire the knowledge from various experts, and store that knowledge into the knowledge base.
Explanation System
  • Traces responsibility and justify the behavior of expert system by firing questions and answers
  • such as Why, How, What, Where, When, Who.
User Interface
  • It allows users to insert their queries with using own Natural Language Processing otherwise menus & graphics.
Knowledge Engeer
Main objective of this engineer is to design system for specific problem domain with using of expert system shell.
System Engineer
To design user interface and declarative format of knowledge base as well as to build inference engine
Users
They are non expert person who want to seek direct advice.
 
Resources
 
Applications
mnemonic : BFAME
Business
Marketing – It deal to large scale company to monitor discounts, quantity and other factors relating to sale.
Accounting – It allowing you to make financial decisions and credit authorization decisions.
Business – This allowing you to operate your business and trainees and evaluate their performance.
Finance
  • Expert system is used in the financial sector for solving these problems, such as –
  • For financial analysis of firms
  • For analyzing the causes of successful or unsuccessful business development
  • For market analysis
Agriculture
Crop Management – This expert system allows farmer for taking good decision about growing specific crop, such as rice, wheat, cucumber etc.
Production Planning – It helps to make profitable cropping plans.
Pest Management – This expert system deals to control pest/weed which are dangerous for cops.
Diagnostic Systems – It helps to manage different types of disease and infections to several crops and plants.
Conservation System – This expert system allows conservation of soil or water as well as it is used for management in agriculture.
Process Control Systems – In this expert system, to implement various sensors for monitoring different types of agricultural activities.
Marketing Advisory Systems – This expert system delivers good advice to farmers to get profits in the agricultural production.
Medical
MYCIN – It helps to diagnose infectious diseases in the blood and prescribing the appropriate antibiotics.
PUFF – To diagnose lung diseases
CADUCEUS – It is used for diagnosing of internal medicine.
Education
Computer Aided Instruction (CAI) – This expert system deals with expert heuristics for solving problems. It delivers for students for learning, and it uses games, puzzles, colorful graphs and sounds to keep the student attentive until the end of session.
Intelligent Tutoring System (ITS) – This expert system is used for learning students into pictorial form. It is totally based on the fuzzy logic, and it delivers best flexibility for representing information as well as responding to individual candidate as per need.
Intelligent Pascal Tutoring System (IPTS) – This expert system is able to keep a student database for identify individual performance and learning ability.
Advantages
  • These systems are highly reproducible.
  • They can be used for risky places where the human presence is not safe.
  • Error possibilities are less if the KB contains correct knowledge.
  • The performance of these systems remains steady as it is not affected by emotions, tension, or fatigue.
  • They provide a very high speed to respond to a particular query.
Limitations
  • The response of the expert system may get wrong if the knowledge base contains the wrong information.
  • Like a human being, it cannot produce a creative output for different scenarios.
  • Its maintenance and development costs are very high.
  • Knowledge acquisition for designing is much difficult.
  • For each domain, we require a specific ES, which is one of the big limitations.
  • It cannot learn from itself and hence requires manual updates.
[TODO] Expert System Vs Traditional System
Uncertainity

Types of Knowledge

Certain
  • Techniques such First-order logic and prepositional logic are used
  • This works due to knowledge being certain
  • Probability value is either 0 or 1, cannot be in between so no uncertainity
  • Uncertain
    • Here we are not sure about the predicates
    • A —> B means B is true iff A is true, but in this uncertain situation we are not sure if A or B is true or false
    • This begs the need of Probabilistic Reasoning
 

Causes of Uncertainity

  • Information taken from unreliable sources
  • Experimental Errors
  • Equipment Fault
  • Temprature Variation
 

Probabilistic Reasoning

  • Way of Knowledge Representation in which we apply concept of probability to indicate uncertainity in knowledge
  • Real world lots of scenarios depend on this
 

Need of Probabilistic Reasoning

  • There are unpredictable outcomes
  • Possibilities of predicates become too large to handle
 

Solve problems with Uncertain Knowledge

Probability Basics

Direct from PPT

Probability

  • Probability can be defined as a chance that an uncertain event will occur.
  • It is the numerical measure of the likelihood that an event will occur.
  • The value of probability always remains between 0 and 1 that represent ideal uncertainties.

Event

Each possible outcome of a variable is called an event

Sample Space

The collection of all possible events is called sample space

Random Variables

Random variables are used to represent the events and objects in the real world.

Prior Probability

The prior probability of an event is probability computed before observing new information.

Posterior Probability

The probability that is calculated after all evidence or information has taken into account. It is a combination of prior probability and new information

Conditional Probability

probability of occurring an event when another event has already happened
Bayes Rule
  • Determines the probability of an event with uncertain knowledge
  • Relates the conditional probability and marginal probabilities of two random events.
  • Bayesian inference is an application of Bayes' Theorem
  • Bayes' theorem allows updating the probability prediction of an event by observing new information of the real world.
  • If cancer corresponds to one's age then by using Bayes' theorem, we can determine the probability of cancer more accurately with the help of age.
 

Bayesian Network, Inference etc.

 

Videos to understand concepts

  • Listening to first 3 videos on 4x speed highly suggested to understand this concept

PPT covering topics

 
Bayesian Networks
  • Probabilistic Graphical model representing set of variables and their conditional dependencies
  • They are probabilitic because they are built from a probability distribution and use probability theory for prediction and anomaly detection
  • Bayesian Neworks can be used for building models form data.
  • The Bayesian Networks consist of Two Parts
    • Directed Acyclic Graph
      notion image
      Table of Conditional Probabilities
      • Give probabilities for value of random variables
      • Each row must sum to 1
      • All variables are boolean
      • Table for node with k-parents will have 2^k rows
Inference
  • Process of inferring unknown properties of a system given already calculated observations

Types of Inferences

Inference By Enumeration
  • By listing all variables
Inference By Variable Elimination
  • By variable removal
Fuzzy Logic
 
Must Watch Video Resources
Watch from 7th to 24th video in playlist. Each video approx 4-5 min.
 
OR
 
Watch Gate smashers
 
  • Superset of Boolean logic
  • Represent with uncertainity
  • Represent with degree
  • Represent belongingness of a member of a crisp set to fuzzy set

Implementation

  • Can be implemented in Software and Hardware of various size
  • Can also be implemented in Micro controllers

Applications

  • It can control machines and consumer products
  • It may not give accurate reasoning, but acceptable reasoning
  • Helps dealing with uncertainity in engineering

Fuzzy Set

  • Collection of values which exist between 0 and 1

Operations

  • Union
    • When you find same value in both sets, include the one with max membership function
  • Intersection
    • When you find same value in both sets, include the one with min membership function
  • Complement Operation
    • ( 1 - membership function ) for whole set
       
Fuzzy Logic Architecture
notion image
 
Fuzzification
  • Converts Crisp input to fuzzy input
  • Divides input signals into five states
      1. Large Positive
      1. Medium Positive
      1. Small
      1. Medium Negative
      1. Large Negative
Rule Base
  • Contains all Rules and IF-ELSE conditions offered by experts to control decision making systems
Inference Engine
  • Degree of match between fuzzy input and the rules. Based on the % match
  • Determines which rules to implement
Defuzzification
Converts fuzzy output to Crisp output
 
Fuzzy Logic Algorithm
1) Initialization process: ▪ Define the linguistic variables. ▪ Construct the fuzzy logic membership functions that define the meaning or values of the input and output terms used in the rules. ▪ Construct the rule base (Break down the control problem into a series of IF X AND Y, THEN Z rules based on the fuzzy logic rules). 2) Convert crisp input data to fuzzy values using the membership functions (fuzzification). 3) Evaluate the rules in the rule base (inference). 4) Combine the results of each rule (inference). 5) Convert the output data to non-fuzzy values (defuzzification).
 
Examples and Membership function (theory)
 
Decision Making
 
 

Utility Theory

  • An agent's preferences over possible outcomes can be captured by a function that maps these outcomes to a real number; the higher the number the more that agent likes that outcome. This function is called a utility function
  • Once you have these numbers, then you can mix them in with probabilities and talk about expected utilities, optimal strategies, discounted future rewards, and many other fun things.
UNIT 4
Machine Learning and it's Paradigms

What is Learning ?

  • Change in system which enables it to perform same task more efficiently the next time
  • Constructing/Modifying representations of what is being experienced
  • Making useful changes in our minds

What is Machine Learning ?

  • First introduced by Arthur Sameul in 1959
  • It is a subset of Artificial Intelligence mainly concerned with development of algorithms
  • Automatically learn from data and past experiences, predict things without explicitly programmed to do so
  • Sample historical data (training data) is fed into these algorithms to produce a mathematical model
  • Brings Computer Science and Statistics together to create predictive models

The Three Paradigms in Machine Learning

notion image
Flow Chart with Brief Overview
 
Supervised Learning
  • Machine learning method in which we provide sample labelled data to the system in order to train it, on the basis of which it predicts the output
  • System creates a model using labelled data to understand the datasets
  • Goal is to map input data with output data
  • Example : Spam filtering
Classification
  • Process of categorizing data into classes (often referred as target,label or categories)
  • Can be performed on both structured and unstructured data
  • Example : Heart disease detection ( binary classification )

Algorithms

  • Decision Tree Algorithm
  • Support Vector Machines
  • Naive Bayes Classification
  • K Nearest Neighbour
  • Artificial Neural Networks
Regression
  • Algorithms attempting to map input variable (x) to numerical or continous output variables (y)
  • Regression prediction problems are usually quantities and sizes since output can be integer as well as float
  • Example : Predict house price from dataset of houses

Algorithms

  • Linear Regression
  • Logistic Regression
  • Regression Decision Tree
  • Artificial Neural Netorks
Unsupervised Learning
  • ML technique in which models themselves find hidden patterns and insights from given data
  • Models are trained using unlabelled data and are allowed to act on that data without any supervision
  • Cannot be directly applied to Classification/Regression since theres no output data to compare
  • Example : Cat and Dog
Clustering
  • Method of grouping objects into clusters such that object with most similarity remains in a group (cat,dog)

Algorithms

  • K - means Clustering
  • Hierarchial Clustering
Association
  • Used to find relationship between variables in a large database. Ex: People who buy X also buy Y
Inductive Learning
  • Knowledge extraction method for large amounts of data since expert system might not be feasible
  • Process where Learner discovers rules by observing examples to find patterns
  • System tries to make a general rule
 
Decision Trees
  • Supervised learning Technique
  • Can be used for Classification and regression but preferred for former
  • Internal Nodes : Features of Dataset
  • Branches : Decision Rules
  • Leaf Nodes : Represents outcome
  • Types of Nodes
    • Decision Node (Used to make any decision)
    • Leaf Node (Output)
    •  
Types of Decision Trees
  • Classification Trees ( Binary Yes / No )
    • Decision Variable is Categorical
  • Regression Trees ( Conintuous Data Types )
    • Oucome variable is continuous
 
  • Best Algorithm to construct a Decision Tree is the ID3 Algorithm
  • ID3 = Iterative Dichotomiser 3
  • It's a Classification Algo following greedy approach of building a decision tree by selecting a best attribute that yields Maximum Information Gain (IG) or Minimum Entropy (H)
  • Entropy is the measure of impurity, disorder or uncertainity
  • Entropy controls how Decision tree decides to split the data
  • Information gain measures how much info a feature gives us about the class
  • Information gain is the main key that is used by Decision Tree Algorithms to construct a Decision Tree.
  • Decision Trees algorithm always tries to maximize Information gain.
  • An attribute with highest Information gain will tested/split first
 

Problems and Formulae[TODO]

 

Complete Decision Tree using ID3 Problem in One Video

Solving this should be enough
 

Sir PPT

 
Resources
Artificial Neural Networks
  • Machine Learning approach that models Human Brain using artificial neurons
  • ANNs have fewer connections compared to human brain
  • Each Neuron receives a number of inputs
  • An activation function is applied to these inputs which result in activation level of neuron
  • Knowledge about the learning task is given in form of training data
  • ANN
    • Neuron Model - Information Processing Unit of the NN
    • Architecture - Links connecting Nerons, each link has an associated weight
    • Learning Algorithm - Training NN by modifying weights in order to model a learning task on training examples
    •  
Neuron and Neural Networks
  • Basic information processing unit of NN
  • Set of links with weights w1,w2,w3
  • Adder function to compute weighted sum of the inputs
  • Activation Function for limiting amplitude of neuron output
  • Bias is external parameter of neuron and can be modelled using extra input
 

Neuron, Activation Functions and BackPropagation Theory

Single Layer Feed-Forward Network
 
Theory from PPT
 
Multi-Layer Feed-Forward Network
Theory from PPT
 
 
 
Reinforcement Learning
  • Reinforcement learning is an area of ML concerned with how intelligent agents can take actions in an environment in order to maximize the reward
  • Training Data does have not answers the agent decides what to do to perform the given task
  • Learns from experience
  • Trained on Reward and Punishment Mechanism
 

PPT Theory

UNIT 5
Natural Language Processing (NLP)
  • Natural language processing (NLP) is a subfield of linguistics, computer science, and artificial intelligence
  • It is concerned with the interactions between computers and human language, in particular how to program computers to process and analyze large amounts of natural language data.
  • The goal is a computer capable of "understanding" the contents of documents, including the contextual nuances of the language within them.
  • The technology can then accurately extract information and insights contained in the documents as well as categorize and organize the documents themselves.
 
notion image
 
History of NLP
1950- NLP started when Alan Turing published an article called "Machine and Intelligence."
1950- Attempts to automate translation between Russian and English
1960- The work of Chomsky and others on formal language theory and generative syntax
1990- Probabilistic and data-driven models had become quite standard
2000- A Large amount of spoken and textual data become available
Phases of NLP
Five main Components of Natural Language processing in AI are:
Morphological and Lexical Analysis
  • Lexical analysis is a vocabulary that includes its words and expressions. It depicts analyzing, identifying and description of the structure of words. It includes dividing a text into paragraphs, words and the sentences
  • Individual words are analyzed into their components, and nonword tokens such as punctuations are separated from the words.
Syntactic Analysis
  • The words are commonly accepted as being the smallest units of syntax. The syntax refers to the principles and rules that govern the sentence structure of any individual languages.
  • Syntax focus about the proper ordering of words which can affect its meaning. This involves analysis of the words in a sentence by following the grammatical structure of the sentence. The words are transformed into the structure to show hows the word are related to each other.
Semantic Analysis
  • Semantic Analysis is a structure created by the syntactic analyzer which assigns meanings. This component transfers linear sequences of words into structures. It shows how the words are associated with each other.
  • Semantics focuses only on the literal meaning of words, phrases, and sentences. This only abstracts the dictionary meaning or the real meaning from the given context. The structures assigned by the syntactic analyzer always have assigned meaning
  • E.g.. "colorless green idea." This would be rejected by the Symantec analysis as colorless Here; green doesn't make any sense.
 
Discourse Integration
  • It means a sense of the context. The meaning of any single sentence which depends upon that sentences. It also considers the meaning of the following sentence.
  • For example, the word "that" in the sentence "He wanted that" depends upon the prior discourse context.
Pragmatic Analysis
  • Pragmatic Analysis deals with the overall communicative and social content and its effect on interpretation.
  • It means abstracting or deriving the meaningful use of language in situations. In this analysis, the main focus always on what was said in reinterpreted on what is meant.
  • Pragmatic analysis helps users to discover this intended effect by applying a set of rules that characterize cooperative dialogues.
  • E.g., "close the window?" should be interpreted as a request instead of an order.
 
notion image
 
Components of NLP
  • There are two components of NLP
Natural Language Understanding
  • Natural Language Understanding (NLU) helps the machine to understand and analyze human language by extracting the metadata from content such as concepts, entities, keywords, emotion, relations, and semantic roles.
  • NLU mainly used in Business applications to understand the customer's problem in both spoken and written language.
  • NLU involves the following tasks -
    • It is used to map the given input into useful representation.
    • It is used to analyze different aspects of the language.
Natural Language Generation
  • Natural Language Generation (NLG) acts as a translator that converts the computerized data into natural language representation.
  • It mainly involves Text planning, Sentence planning, and Text Realization.
  • An NLG System contains
    • Discourse Planner
      • What will be generated, which sentences
    • Surface Realizer
      • Realizes a sentence from its internal representation
    • Lexical Selection
      • Selecting the correct words describing the concepts
💡
NLU is more difficult tha NLG
notion image
Applications of NLP
  • Question Answering
  • Spam Filtering
  • Sentiment Learning
  • Machine Translation
  • Spelling Correction
  • Speech Recognition
  • Chatbot
  • Information Extraction
  • Natural Language Understanding (NLU)
Building an NLP Pipeline
Sentence Segmentation
Sentence Segment is the first step for building the NLP pipeline. It breaks the paragraph into separate sentences.
Example: Consider the following paragraph -
Independence Day is one of the important festivals for every Indian citizen. It is celebrated on the 15th of August each year ever since India got independence from the British rule. The day celebrates independence in the true sense.
Sentence Segment produces the following result:
  1. "Independence Day is one of the important festivals for every Indian citizen."
  1. "It is celebrated on the 15th of August each year ever since India got independence from the British rule."
  1. "This day celebrates independence in the true sense."
Word Tokenization
JavaTpoint offers Corporate Training, Summer Training, Online Training, and Winter Training.
Word Tokenizer generates the following result:
"JavaTpoint", "offers", "Corporate", "Training", "Summer", "Training", "Online", "Training", "and", "Winter", "Training", "."
Stemming
  • Stemming is used to normalize words into its base form or root form. For example, celebrates, celebrated and celebrating, all these words are originated with a single root word "celebrate."
  • The big problem with stemming is that sometimes it produces the root word which may not have any meaning.
  • For Example, intelligence, intelligent, and intelligently, all these words are originated with a single root word "intelligen." In English, the word "intelligen" do not have any meaning.
Lemmatization
  • Lemmatization is quite similar to the Stemming. It is used to group different inflected forms of the word, called Lemma.
  • The main difference between Stemming and lemmatization is that it produces the root word, which has a meaning.
Identifying Stop Words
  • In English, there are a lot of words that appear very frequently like "is", "and", "the", and "a". NLP pipelines will flag these words as stop words.
  • Stop words might be filtered out before doing any statistical analysis.
  • Example: He is a good boy.
Dependancy Parsing
Dependency Parsing is used to find that how all the words in the sentence are related to each other.
POS Tags
POS stands for parts of speech, which includes Noun, verb, adverb, and Adjective. It indicates that how a word functions with its meaning as well as grammatically within the sentences. A word has one or more parts of speech based on the context in which it is used.
Example: "Google" something on the Internet.
In the above example, Google is used as a verb, although it is a proper noun.
Named Entity Recognition
  • Named Entity Recognition (NER) is the process of detecting the named entity such as person name, movie name, organization name, or location.
  • Example: Steve Jobs introduced iPhone at the Macworld Conference in San Francisco, California.
Chunking
Chunking is used to collect the individual piece of information and grouping them into bigger pieces of sentences.
What makes NLP Difficult ?
Ambiguity
  • Lexical Ambiguity
  • Syntactic Ambiguity
  • Referential Ambiguity
Grammars and Parsing
Theory from PPT
Some Extra Stuff
NLP APIs
  • IBM WATSON API
  • CHATBOT API
  • SPEECH TO TEXT API
  • GOOGLE CLOUD NATURAL LANGUAGE API
NLP Libraries
  • Scikit-learn
  • Natural language Toolkit (NLTK)
  • Pattern
  • TextBlob
  • Quepy
  • SpaCy
  • Gensim
Resources
 
Automatic Speech Recognition

Automatic Speech Recognition Software (ASR)

  • Automatic speech recognition (ASR) is the use of computer hardware and software-based techniques to identify and process human voice. It is used to identify the words a person has spoken or to authenticate the identity of the person speaking into the system.
  • Automatic speech recognition is also known as automatic voice recognition (AVR), voice-to-text or simply speech recognition
Easy to understand Infographic
notion image
Training of ASR
Human Tuning
This is a relatively simple means of performing ASR training.
It involves human programmers going through the conversation logs of a given ASR software interface and looking at the commonly used words that it had to hear but which it does not have in its pre-programmed vocabulary.
Those words are then added to the software so that it can expand its comprehension of speech.
Active Learning
Active learning is the much more sophisticated variant of ASR and is particularly being tried with NLP versions of speech recognition technology.
With active learning, the software itself is programmed to autonomously learn, retain and adopt new words, thus constantly expanding its vocabulary as it’s exposed to new ways of speaking and saying things.
This, at least in theory, allows the software to pick up on the more specific speech habits of particular users so that it can communicate better with them.
So for example, if a given human user keeps negating the autocorrect on a specific word, the NLP software eventually learns to recognize that particular person’s different use of that word as the “correct” version.
Famous ASR Systems
DRAGON
  • Dragon Dictate launched in 1990 for a staggering $9,000, touting a dictionary of 80,000 words and features like natural language processing (see the Infoworld article above).
  • These tools were time-consuming (the article claims otherwise, but Dragon became known for prompting users to ‘train’ the dictation software to their own voice). And it required that users speak in a stilted manner: Dragon could initially recognize only 30–40 words a minute; people typically talk around four times faster than that.
  • But it worked well enough for Dragon to grow into a business with hundreds of employees, and customers spanning healthcare, law, and more. By 1997 the company introduced Dragon NaturallySpeaking, which could capture words at a more fluid pace — and, at $150, a much lower price-tag⁸.
HARPY
  • the U.S. Department of Defense’s ARPA (the agency now known as DARPA) funded a five-year program called Speech Understanding Research.
  • This led to the creation of several new ASR systems, the most successful of which was Carnegie Mellon University’s Harpy, which could recognize just over 1000 words by 1976.
Resources
Machine Vision
Machine Vision Hierarchy
  • Low-level vision: includes process image for feature extraction.
  • Intermediate-level vision: includes object recognition and 3D scene Interpretation
  • High-level vision: includes conceptual description of a scene like activity, intention and behavior.
Machine Vision Techniques
  • Image Classification
  • Object Detection
  • Object Tracking
  • Semantic Segmentation
  • Instance Segmentation
Machine Vision Applications
  1. Robotics
  1. Medicine
  1. Security
  1. Transportation
  1. Industrial Automation
Resources
AI Today and Tomorrow
AI Achievements
[TODO] to further explain
  • Text Generation
  • AI-induced Face Recognition
  • Advances in Algorithm Design
  • Drug Discovery
  • Speech Recognition
  • Lung Cancer Diagnosing
  • COVID 19 detection through X-Rays
Ubiquitous AI
  • Personal AI
  • Organs Growing on Microchips
  • Autonomous Vehicles
  • Blockchain
  • 2D Material
  • IOT