5 - Machine learning: Trends, perspectives, and prospects - Jordan & Mitchell
(2015)
Paper over machine learning
• Wat is machine learning?
• Hoe verschillen supervised learning, unsupervised learning en reinforcement learning
van elkaar?
Abstract:
- machine learning addresses the following questions:
how to build computers that improve automatically through experience
what are the fundamental statistical-computational-information-theoretic laws that
govern all learning systems, including computers, humans and organizations?
- machine learning: train a system by showing it examples of desired input-output behavior
- learning problem: the problem of improving some measure or performance when executing
some task, through some type of training experience
- machine learning algorithms: searching through a large space of candidate programs, guided
by training experience, to find a program that optimizes the performance metric
→ they vary greatly in two ways:
- how they represent candidate programs (decision trees, mathematical functions etc.)
- how they search through this space of programs (optimization algorithms etc.)
- algorithms that focus on function approximation problems: task is embodied in a function and
the learning problem is to improve the accuracy of that function, with experience consisting of a
sample of known input-output pairs of the function
-goal of algorithm: theoretically characterize the capabilities of specific learning algorithms and
the inherent difficulty of any given learning problem
- sample complexity: how much data are required to learn accurately
- computational complexity: how much computation is required
- mobile devices and embedded computing permit large amounts of data to be gathered about
individual humans, and machine-learning algorithms can learn form these data to customize
their services to the needs and circumstances of each individual
→ can be connected, so that an overall service emerges from the data of many people
- supervised learning systems: exemplify the function approximation problem, where the
training data take the form of a collection of (x, y) pairs and the goal is to produce a prediction
y* in response to a query x*
(2015)
Paper over machine learning
• Wat is machine learning?
• Hoe verschillen supervised learning, unsupervised learning en reinforcement learning
van elkaar?
Abstract:
- machine learning addresses the following questions:
how to build computers that improve automatically through experience
what are the fundamental statistical-computational-information-theoretic laws that
govern all learning systems, including computers, humans and organizations?
- machine learning: train a system by showing it examples of desired input-output behavior
- learning problem: the problem of improving some measure or performance when executing
some task, through some type of training experience
- machine learning algorithms: searching through a large space of candidate programs, guided
by training experience, to find a program that optimizes the performance metric
→ they vary greatly in two ways:
- how they represent candidate programs (decision trees, mathematical functions etc.)
- how they search through this space of programs (optimization algorithms etc.)
- algorithms that focus on function approximation problems: task is embodied in a function and
the learning problem is to improve the accuracy of that function, with experience consisting of a
sample of known input-output pairs of the function
-goal of algorithm: theoretically characterize the capabilities of specific learning algorithms and
the inherent difficulty of any given learning problem
- sample complexity: how much data are required to learn accurately
- computational complexity: how much computation is required
- mobile devices and embedded computing permit large amounts of data to be gathered about
individual humans, and machine-learning algorithms can learn form these data to customize
their services to the needs and circumstances of each individual
→ can be connected, so that an overall service emerges from the data of many people
- supervised learning systems: exemplify the function approximation problem, where the
training data take the form of a collection of (x, y) pairs and the goal is to produce a prediction
y* in response to a query x*
