A subfield of AI that comprises the study of algorithms which have a capability to improve themselves automatically through experience to solve problems without external instructions, by using previously trained models.
1.5.3.
Intelligent agents
Autonomous entity, which acts, directing its activity toward achieving goals, upon an environment using observation through sensors and consequent actuators.
3.6.5.4
Deep learning
Machine learning method based on characterization of data learning.
1.5.3
Reinforcement learning
Algorithms, in which an agent decides what to do to perform the given task to maximize the given function.
1.5.7
Shallow learning
Techniques that separate the process of feature extraction from learning itself.
3.6.5.1
Supervised learning
Algorithms, which develop a mathematical model from the input data and known desired outputs.
1.5.7
Alg. 1.1.
Unsupervised learning
Algorithms, which take a set of data consisting only of inputs and then they attempt to cluster the data objects based on the similarities or dissimilarities in them.
1.5.7.
Alg. 1.2.
Decision tree
Tree‐structure resembling a flowchart, where every node represents a test to an attribute, each branch represents the possible outcomes of that test, and the leaves represent the class labels.
J48
Open source Java implementation of the C4.5 algorithm that builds decision trees from a set of training data using the concept of information entropy.
6.6.4
Genetic/evolutionary algorithms
Set of evolutionary algorithms, which take an inspiration from genetic evolution theories.
3.6.4, 3.6.5.4
Alg. 1.3
Hidden Markov models
Algorithm that builds up a set of states producing outputs with different probabilities with the goal to find out the sequence of states that results in the observed outputs.
K‐means
Clustering algorithm that uses a distance function to distribute all data pieces between k clusters defined by their centroid position in the feature space.
3.6.2
K‐nearest neighbor
Classification algorithm that uses a distance function in order to determine to which class to assign the new element by finding K closest elements in the feature space.
3.6.3, 5.3.5.4
Naive Bayes
Algorithm that consists of applying the Bayes theorem in order to find a distribution of conditional probabilities among class labels, with the assumption of independence between features.
Random forest
An ensemble learning method that builds a large group of independent decision trees, and outputs the mode of the label predictions of all the trees.
6.6.4
Sec.6.6.4
Support vector machine
Binary classification algorithm that creates a hyper plane that separates the data into two classes with the objective to maximize the gap perpendicular to the plane, allowing better generalization.
Please note: I realize that there exist various definitions and even understandings of these terms’ meaning. I have chosen to follow up the definitions given in the publications of the NIST Computer Security Resource Center (see https://csrc.nist.gov/glossary), first (see Section I.6) and then proceed with others (see Section I.7). Even those publications are ambiguous in some cases and provide different meanings too. I have chosen ones, which are followed up in this book. I do not intend to make this list all inclusive or exclusive.
NIST SP 800‐53 Rev. 5 CNSSI 4009 Security and Privacy Controls for Information Systems and Organizations, September 2020, available at doi.org/10.6028/NIST.SP.800‐53r5
NIST 800‐63 Digital Identity Guidelines, June 2017, available at СКАЧАТЬ
