0% found this document useful (0 votes)
56 views

Artificial Intelligence

Artificial intelligence is the study of programming computers to exhibit intelligent behavior. The document discusses several topics in AI including learning systems, knowledge representation and reasoning, planning, knowledge acquisition, intelligent search, logic programming, soft computing, and management of imprecision and uncertainty. It also outlines different branches of AI such as symbolic AI, statistical AI, and computational intelligence.

Uploaded by

alwinhilton58
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
56 views

Artificial Intelligence

Artificial intelligence is the study of programming computers to exhibit intelligent behavior. The document discusses several topics in AI including learning systems, knowledge representation and reasoning, planning, knowledge acquisition, intelligent search, logic programming, soft computing, and management of imprecision and uncertainty. It also outlines different branches of AI such as symbolic AI, statistical AI, and computational intelligence.

Uploaded by

alwinhilton58
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

UNIT -1

ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING


DEFINITION OF AI:
Artificial intelligence is a branch of computer science that studies how to
program computers to exhibit apparently intelligent behaviour.
Father of ‘AI’ – John McCarthy
Turing test for ‘ML’ – Alan Turing
TOPICS OF AI:
Learning systems:
1. *Definition of Learning: Herbert Simon defines learning as any process by
which a system improves its performance through experience.
2. *Learning Example: Using a child learning pronunciation as an example,
where the system adjusts based on error signals until the error rate is negligible.
3. *Adaptive Learning: Describes the process as adaptive learning, continuously
adjusting parameters to achieve results close to the training dataset.
4. *Supervised Learning: Highlights the importance of Artificial Neural
Networks (ANN) in solving supervised learning problems.
5. *Other Learning Methods: Mentions inductive learning (e.g., 'birds fly') and
analogy-based learning (e.g., comparing electron motion to planetary
movement).
6. *Overall: Learning involves continuous adjustment of parameters, with
various methods such as supervised learning and analogy-based learning
contributing to AI systems.
Knowledge representation and reasoning:
1. *Definition of Knowledge: Expertise and skills acquired by individuals,
challenging to represent in computers.
2. *Knowledge Base in AI: Analogous to a database, crucial for building expert
systems and facilitating reasoning.
3. *Efficiency in Reasoning: Achieving a predefined target state with fewer
transitions enhances the efficiency of reasoning systems.
4. *Minimization of Intermediate States: Efficiency requires minimizing
intermediate states, emphasizing the need for an organized and complete
knowledge base.
5. *Importance of Knowledge Organization: The organization of knowledge is
crucial in knowledge engineering for effective problem-solving.
6. *Knowledge Representation Techniques: Various techniques in AI, including
production rules, predicate logic, frames, filler and slots, semantic nets, are used
for knowledge representation.
7. *Choice of Representation: The selection of knowledge representation
depends on user preference and the nature of applications.
8. *Reasoning Problem Focus: Testing the satisfiability of a goal from a given
set of data and knowledge is the primary concern in reasoning problems.
Planning:
1. *Planning in AI: Significant area in artificial intelligence with a distinction
from reasoning.
2. *Problem Planning Definition: Involves determining a methodology to
achieve a successful result from known initial states.
3. *Applications: Extensively used in robotics and navigational problems,
contributing to automated planning.
4. *Focus on Decomposition: Planning emphasizes breaking down an original
problem into suitable subunits and interacting on these subunits during problem-
solving.
5. *Components of Planning System: Major components include heuristic
functions, a set of rules in the problem state, and a deduction process.
6. *Deduction Process: Aims to find a solution by identifying a sequence of
operators that transition an initial state to the goal state.
Knowledge Acquisition:
1. *Knowledge Acquisition Significance: Challenging for both humans and
computers.
2. *Scope of Knowledge Acquisition: Involves creating new knowledge,
activating structures in existing knowledge, learning from the environment, and
fine-tuning.
3. *Automated Acquisition: Machine learning enables automated knowledge
acquisition, a dynamic research area in AI.
4. *Knowledge Storage Mechanisms: Various mechanisms like frames,
conceptual dependency, scripts, and semantic nets are used to store acquired
knowledge.
5. *Data Structures for Knowledge Store: Frames, conceptual dependency,
scripts, semantic nets are data structures aligned with different knowledge
representation schemes.
Intelligent search:
1. *Intelligent Search in AI: Search problems in artificial intelligence are non-
deterministic, dependent on given datasets.
2. *Deterministic vs. Non-deterministic: Unlike deterministic computer science
searches, AI searches lack a predefined order for visiting elements in the search
space.
3. *Search Techniques: Various techniques include Best-first search, greedy
search, BFS, DFS, among others.
4. *Mining Databases: Choice of search technique depends on the usage and
nature of data in database mining.
Logic programming:
1. *Evolution of Logic Symbols: Logic symbols have evolved over the past
century for diverse uses.
2. *Mathematical Tools: Mathcautie is used for designing tools to represent
logical statements with symbolic operators.
3. *Proportional Logic: Deals with binary statements using Boolean operators to
handle complex real-world situations.
4. *Predicate Logic: Also known as first-order logic, used in AI-based
programming; PROLOG (PROgramming in LOGic) is a classic example.
5. *Logic Programming: A significant area in AI research, covering knowledge
representation structures.
6. *PROLOG: Widely used for artificial intelligence programming, aims to
expand to handle temporal models and parallel programming environments.
7. *LISP: Another AI programming language (LISt Processor) capable of
handling strings but surpassed by modern languages like R, Python, and Java.
8. *Modern Languages: Present languages like R, Python, and Java outperform
older languages, offering various supporting packages for machine learning.
Open-source machine learning benchmarking tools are also available.
Soft computing:
1. *Soft Computing Overview:
- Problem-solving technique inspired by nature.
- Utilizes statistical learning, fuzzy logic, genetic algorithms, fractals,
probabilistic reasoning, and chaos theories.
2. *Nature-Inspired Techniques:
- Soft computing techniques mimic the intelligence found in nature.
- Draws inspiration from the human brain's ability to process data, learn,
generalize from rules, recognize patterns, and make decisions.
3. *Engineering Model:
- Human brain serves as an engineering model, demonstrating tasks using
heuristic approximations, low precision, or less generality based on the problem
at hand.
4. *Transfer of Abilities:
- Goal is to transfer human-like abilities into designed models, algorithms,
machines, and AI systems.
- Aims to solve tasks with reasonable accuracy, cost, and time based on
previous experience.
5. *Zadeh's Principle (1994):
- Soft computing's guiding principle is to explore tolerance for imprecision,
uncertainty, and partial truth to achieve tractability, robustness, and low solution
cost.
6. *Techniques and Tools:
- Soft computing encompasses artificial neural nets, fuzzy logic, genetic
algorithms, inductive logic programming, belief calculus, etc.
- Tools used individually or jointly depending on the application domain.
Management of imprecision and uncertainty:
1. *Imprecision and Uncertainty in AI:
- Data and knowledge bases in AI problems, like reasoning and planning,
often contain incompleteness.
- Incompleteness leads to imprecision (lack of appropriate data) and
uncertainty (lack of certainty in knowledge).
2. *Imprecision in Data:
- Imprecision occurs due to a lack of appropriate data and poor authenticity of
sources in databases.
3. *Uncertainty in Knowledge:
- Uncertainty arises from incomplete knowledge in the knowledge base.
4. *Challenges in Reasoning:
- Dealing with both imprecision in data and uncertainty in knowledge poses a
complex challenge in reasoning.
5. *Techniques for Handling Incompleteness:
- Various techniques and tools exist for reasoning under incomplete data and
knowledge.
- Examples include fuzzy, stochastic, and belief network models.
6. *Stochastic Systems:
- In stochastic systems, the sum of transitions from one state to another is
strictly unity.
7. *Fuzzy Reasoning:
- Fuzzy reasoning allows the sum of membership values for transitions to be
greater than or equal to one.
- Useful in solving problems involving uncertainty.
8. *Belief Network Model:
- Updates stochastic fuzzy beliefs in the network, stopping when equilibrium
is reached.
9. *Fuzzy Petri Net:
- Applied as a fuzzy tool for handling both imprecision of data and indecision
of knowledge through a unified approach.
BRANCHES OF AI:
Symbolic ai:
1. Symbolic artificial intelligence dominated research from the 1950s to the
1980s.
2. It relied on high-level representations in problem domains.
3. Projects like expert systems aimed for general human-like intelligence.
4. Post-1980s, focus shifted to specific subproblems.
5. Doubts emerged about symbolic systems imitating all human processes.
6. Challenges included limitations in mimicking robotics and perception
learning.
Statistical ai:
1. Statistical AI draws inspiration from mathematics for diverse applications.
2. Inclusion of statistical approaches enhances learning in AI research.
3. Computational models based on mathematical principles excel in real-world
applications.
Computational ai:
1. *Computational Intelligence:
- Solves real-world problems using soft computing techniques.
- Includes artificial neural networks, support vector machines, fuzzy logic,
heuristic search, etc.
- Modifies statistical techniques like the Markov model for machine learning.
2. *Branches of AI (John McCarthy):
1. Logical AI: Represents world knowledge in mathematical logical language.
2. Search AI: Examines possibilities efficiently, crucial for intelligent
applications.
3. Pattern Recognition: Uses algorithms for detecting patterns in various data.
4. Representation: Facts about the world represented in symbolic form, used
in AI algorithms.
5. Inference: Derives logical conclusions; involves non-monotonic reasoning.
6. Common Sense Knowledge and Reasoning: Focuses on human-like
understanding but faces challenges.
7. Learning from Experience: AI programs can learn from specialized
applications but have limitations.
8. Planning: Deals with world facts and action effects, crucial for goal
achievement.
9. Epistemology: Studies the philosophy of knowledge in AI.
10. Ontology: Examines the types of things that exist in AI programs.
11. Heuristics: Uses mathematical functions to discover ideas embedded in
programs.
12. Genetic Programming: Soft computing technique evolving from genetic
algorithms to solve tasks

APPLICATION OF AI:
1. *Game Playing:
- Game theory models interactive situations.
- AI applied successfully in games with large search spaces.
- Chess, a prime example of AI success.
2. *Expert Systems:
- AI software solving specialized problems.
- MYCIN: Early expert system for medical diagnoses.
- Knowledge engineering extracts human expertise for AI decision-making.
3. *Natural Language Processing (NLP):
- Machines understanding and processing human languages.
- Tasks include summarization, language generation, and sentiment analysis.
- Fusion of computer science and linguistics.
4. *Image Understanding:
- AI interprets two-dimensional arrays of pixels in images.
- Low, medium, and high-level vision processes.
- Pattern classification and knowledge-based interpretation used.
5. *Robotics:
- AI applied to robotics for precision tasks and space exploration.
- Deep neural networks enhance robotic capabilities.
- Robotics expected to flourish in the 21st century.
6. *Machine Learning:
- Involves designing algorithms for computers to learn from data.
- Decision tree learning, Bayesian networks, and association learning are
examples.
- Neural networks mimic human brain for learning and recognition.
7. *Inductive Logic Programming (ILP):
- Uses logic programming for rule creation.
- Reasoning from specific observations to create general hypotheses.
- Mainly applied to prediction and classification problems.
8. *Hidden Markov Model (HMM):
- Statistical model used in pattern recognition, speech, and language
processing.
- Consists of hidden states, visible states, and a Markov process.
- Utilizes transition probability matrix and emission probability matrix.
9. *Pattern Recognition:
- Focus on supervised and unsupervised learning.
- Aims to classify data based on knowledge or statistical information.
- Application areas include optical character recognition and speech
processing.

You might also like