What is the msg Machine Learning Catalogue?
This is a catalogue of machine learning methods. It is intended for use when:
The user wishes to obtain a list of methods that would be appropriate for a specific use case.
The user wishes to find out about a specific technique that is mentioned in an article, program or lecture.
Machine Learning Meta Model
- Move the cursor over the filled circles in the graphic model to highlight the relations and dependencies.
- After a short delay a popup with a brief description of the building block appears.
- Move the cursor over the arrow at the end of a line to check which building block references this targeted block.
The catalogue lists:
- Algorithms, defined as normal machine-learning procedures operating on data.
- Supporting techniques, defined as procedures that can each be used to modulate the behaviour of one or more algorithms.
For each algorithm or supporting technique, there is a textual description that aims to provide an informal summary of what the algorithm does. In the textual descriptions, accessibility is deliberately prioritised over rigour; the reader who requires more details will generally have no problems in finding precise mathematical descriptions of the various algorithms elsewhere on the internet.
Both algorithms and supporting techniques specify the following structured information:
- Aliases: alternative names for this algorithm or supporting technique.
- Subtypes: specific versions of this algorithm or supporting technique.
- Functional building blocks: what this algorithm or supporting technique is used for.
- Learning styles: how this algorithm or supporting technique is trained.
- Relevance: whether this algorithm or supporting technique is still in current use.
- Mathematically similar to: other algorithms or supporting techniques that are equivalent or isomorphic to this one.
Algorithms additionally specify the following structured information:
- Input data types: the structure of the data that the algorithm would be fed by a program using it.
- Internal models: internal structures that the algorithm uses for its calculations.
- Output data types: the structure of the data that the algorithm would return to a program using it.
- Parametricities: to what extent the user specifies the model to the algorithm and to what extent the algorithm learns the model itself.
- Uses: other algorithms that this algorithm uses in its operation.
- Sometimes supports: other algorithms whose operation is sometimes assisted using this algorithm.
Supporting techniques additionally specify the following structured information:
- Typically supports: Algorithms that are typically assisted using this supporting technique.