Understanding Knowledge Representation and Artificial Intelligence

Posted on Jun 4, 2025 in Other university degrees and diplomas

Knowledge Representation in AI

Knowledge representation enables AI mechanisms to find efficient solutions. Key questions include: How to choose a formalism for simple translation from the real world? What representation allows for efficient usage?

Information vs. Knowledge

Information refers to basic, uninterpreted data used as input, such as blood test figures or chemical plant sensor data.

Knowledge is structured data modeling experience in a domain or interpretations of basic data. For example, interpreting blood test values or chemical plant sensor data to determine if they are normal, high, low, disturbing, or dangerous.

Knowledge-Based Systems

Knowledge-Based Systems involve computer systems with “intelligence” for problem-solving, automated reasoning, and machine learning.

Sciences Contributing to AI

  • Philosophy: (from 400 BC) conceived of the mind working like a machine, encoding knowledge in an internal language to determine the right action.
  • Mathematics: provided tools for handling certainty and probabilistic logic, paving the way for reasoning algorithms.
  • Psychology: reinforced the idea that humans and animals process information like machines.
  • Linguistics: demonstrated that language use fits this model.
  • Computer Engineering: offered devices to realize AI applications. AI programs rely on advances in computer speed and memory.

AI Techniques and Fields

Key techniques and fields include: Machine Learning, Knowledge Engineering, Fuzzy Logic, Neural Networks, Reactive Systems, Multi-Agent Systems, Rule-Based Systems, Case-Based Reasoning, Expert Systems, and Bayesian Networks.

Perception and Representation

Representing world knowledge with logic, using mechanical methods to deduce new logical knowledge (infer) and new real-world knowledge.

Types of Problems

Problems can be categorized as single-state, multiple-state, or contingency problems.

Logic Completeness

Logic completeness refers to mathematical models that are abstractions of the actual model.

Basic Search Techniques

Effective search techniques should:

  1. Respond flexibly to various situations.
  2. Make sense of ambiguous or contradictory messages.
  3. Recognize the importance of different elements in a given situation.
  4. Find similarities between situations despite differences.
  5. Establish distinctions between situations despite similarities.

Typical AI Problem Solutions

Solutions can be found in two ways:

  1. A state space that satisfies the requirements.
  2. A path specifying how to traverse to find a solution.

Requirements for a Good Search Technique

A good search technique must:

  • Be systematic.
  • Make changes to the database.

Tree vs. Graph

Nodes of the search tree: state + road. The search tree is constructed incrementally and reflects a process on the graph search space of states.

Properties of Breadth-First Search

Breadth-first search may not always complete. When it does, it necessarily obtains a minimum solution. Consider time and space complexity.