Master This You Master Everything Kind Of Flashcards
Data Science is
The science of using data as key part in the process
of creating knowledge.
Artificial Intelligence is…
1) Systems that are (partially) intelligent.
2) The science of engineering technologies that
fulfill some criteria of intelligence.
Intelligence as…
1) Acting humanly? 2) Acting rationally? 3) Thinking
humanly? 4) Thinking rationally?
5) an entity’s capability
▪ to adapt behavior
▪ in response to own interactions with environment
▪ to a changing environment
▪ in order to achieve goals
= an entity’s capability to learn from experience.
What does a system need to be able to do in order
to have a chance at passing as intelligent?
Perceive
Senses and sensors – Digital systems: Data received through interactions with other systems, and with
humans.
Think
“Brain” - Memory, knowledge representation, reasoning. This lecture focused on multiple paradigms of
how to represent knowledge, and how to reason/think given data/knowledge in a particular form
Broadly, two approaches:
▪ Symbolic (rules, object-oriented KR, graphs)
▪ Sub-symbolic, data-driven (graphs, vector representations, neural networks – more from statistics
would be: Bayesion networks, Hidden Markov models, etc.)
Act
Human body, and actuators – Digital systems: Interactions with other systems, and with humans
(interactive systems)
Key capabilities of intelligent systems
What is a representation?
A representation Y conforms in a systematic manner to
X , preserving pre-selected characteristics of X.
A representation always loses something; it is an
approximation.
In this lecture:
▪ Representations of knowledge or facts about the
world
▪ … that a computer uses to think about the (aspects
of) the world.
Knowledge or data? I
General, terminological knowledge vs. information
about concrete entities (~data, facts)
▪ Symbolic knowledge representation formally
represents (human) knowledge.
▪ As a basis for computers to think/reason
▪ Could also have other uses, like communication
between humans.
▪ Vectors representations of complex entities
▪ General knowledge is implicitly encoded in the choice
of which entities to represent, and in the choice of
features (what elements constitute a vector). No formal
representation of general knowledge!
▪ Every single vector is a datum; constitutes a fact.
Different approaches to AI differ in terms of
how and whether they represent knowledge
and data
… and subsequently in how they can “think”.
Symbolic AI: Explicit, formal (different formalisms!)
representation of knowledge, and also of data. Logicbased reasoning is possible; typically high knowledge
engineering effort.
Subsymbolic AI: Knowledge is typically not formally
and explicitly encoded; explicit representation typically
of data; human-level general knowledge tends to be
implicit in many system parameters/system choices.
Choices of how and what knowledge or data to
represent affect fundamentally what kinds of questions
one can ask of the system/how the system can think
Rule-based systems:
▪ Is available knowledge/available data consistent?
▪ Is X true?
▪ What is all general knowledge/all facts available?
Object-oriented knowledge-representations
▪ Is available knowledge/available data consistent?
(depends on formalism)
▪ What is known about X (class, relationship, instance)?
▪ How do X and Y relate to each other?
Choices of how and what knowledge or data to
represent affect fundamentally what kinds of questions
one can ask of the system/how the system can think
Graphs
▪ What is known about X (class, relationship or instance
– depends a bit on semantics given to graph)?
▪ How are X and Y (two nodes) related to each other?
▪ What are central nodes, in-between nodes, subgraphs, … (questions from graph mathematics, need
to be assigned semantics in the domain of
knowledge)?
▪ How can something (information, energy, …) spread
through the network?
Choices of how and what knowledge or data to
represent affect fundamentally what kinds of questions
one can ask of the system/how the system can think
Correlation: What is the relationship between x and y (variables)?
Prediction: Given x, what is the likelihood of y?
Classification: Given training data, can we learn pre-defined
categories/data labels for unseen data?
Clustering: Can data be partitioned into sub-groups?
Other structure identification: Can data be described by a priori
unkown structures (e.g., factor analysis, social network analysis)?
Other mathematical modelling: Does the given data confirm a given
mathematical model? Which model of the phenomenon would explain
the observed data?
Questions data scientists ask of data; questions „AIs“
can be programmed to ask „themselves“
What this lectures covers (very little)
We covered:
▪ What is data science, what is AI, data and knowledge, data
structures and knowledge representation
▪ knowledge representation and reasoning paradigms and
underlying assumption. This achieves „thinking and acting
(in a digital world) rationally)
▪ A very first ML algorithm, which achieves „learning from
experience“
Understand what data science is
▪ Understand what artificial intelligence is
▪ Understand the relationship between data science and
artificial intelligence
▪ and the relationships of knowledge representation and
reasoning approaches taught in this lecture to both.
▪ Remember and understand different definitions of
intelligence, and apply them to concrete examples
▪ Remember and explain building blocks of intellient
systems
