Please enable JavaScript.
Coggle requires JavaScript to display documents.
Artificial Intelligence A technology that enables a machine to simulate…
Artificial Intelligence
A technology that enables a machine to simulate intelligence
Categorisations
Type 1
Categorisation based on capabilities
Narrow AI
Dedicated for one task
General/Human Level AI
Perform like humans
Strong/Super AI
More intelligent than humans
Type 2
Categorisation based on functionality
Reactive Machines
Most basic type, do not store memories or past experiences
Focus on current scenarios and react
IBM Deep Blue, Google AlphaGo
Limited Memory
Stores past experiences or some other data for a short period of time.
e.g. self driving cars storing recent location/speed of nearby cars
Theory of Mind
Should be able to understand human emotions, people, beliefs, social interaction
Not yet developed
Self-Awareness
Super intelligent machines that have their own consciousness, sentiments, and self-awareness
Subsets
Machine Learning
Algorithms that allow a machine to automatically learn from past data without programming explicitly
Learning Problems
Supervised Learning
Using a model to learn a mapping between input examples and the target variable
Operates on input and target output data
Classification
Involves predicting a class label
Regression
Involves predicting a numerical label
Algorithms
Decision Trees
Support Vector Machines
Unsupervised Learning
Using a model to describe or extract relationships in data.
Operates only on input data.
Clustering
Involves finding groups in data. e.g. k-Means
Density Estimation
Involves summarising the distribution of data. e.g. Kernel Density Estimation
Visualisation
Involves creating plots of data. e.g. scatter plot matrix
Projection
Involves creating lower-dimensional representations of data
e.g. principal component analysis
Reinforcement Learning
An agent operates in an environment and learns to operate using feedback.
Algorithms
Q-learning
Temporal-Difference Learning
Deep Reinforcement Learning
Hybrid Learning Problems
Semi-Supervised Learning
Where training data contains very few labelled examples and a large number of unlabeled examples.
Self-Supervised Learning
Framing an unsupervised learning problem as a supervised learning problem in order to apply supervised learning algorithms to solve it.
Auto-encoders
Encode the original input to a smaller dimension then use the original input as the target output
Generative Adversarial Networks (GANs)
Build from a generator network and a discriminator network. The discriminator network attempts to classify whether a given image is from the unlabelled input domain and the generator attempts to create images good enough to fool the discriminator.
Multi-Instance Learning
A supervised learning problem where individual examples are unlabelled; instead, bags or groups of samples are labelled.
Statistical Inference
Natural Language Processing (NLP)
The ability of a computer to understand, analyse, manipulate, and generate human language.
Text Generation
Question Answering
Context Extraction
Classification
Machine Translation