How to fill Jupiter Doc
Minimal requirements for an ipynb file doc
Files have to be named as follows :
- Documentation_AlgoXXX // Documentation_QualityIndexYYY // …
Every doc file should at least contains :
- Title : Name of features description. Ex : K-Means. // QUESTION : Does the title should be space separated or as lib class name : Ex : HardClustering / Hard Clustering
- Labels / Tags :
- Principle : A description as concise and informative as possible of :
- What the algorithm does (to specify).
- What kind of data it takes in input.
- What kind of model it returns.
- Scalability.
- Complexity.
- Computing : In $O(n)$, $O(n^2)$, $O(n.k)$, …
- Memory (optional if not easily accessible).
- Complexity.
- Input : The type of input data.
- Parameters.
- 1 : parameter 1 + description
- 2 : parameter 2 + description
- .
- .
- Output.
- Start by exposing the output model type, for ex : KMeansModel / Clusters / HardClustering.
- Describe its features and meaning.
- Associated visualization : List of visualization associate to pipeline output.
- Practical strategies
- Recommended association
- Business case
- Usage
Where each bullet point have to be in its own jupyter notebook cell.
Predictor
- Input
- Predictor hyperparameters
- Queries type
- Single query
- Collection type
- Output
- Single query
- Multiple queries
Clustering Algorithms
- Predictor : the default predictor associate to this output model.
- IFF the output is not a Clusters or a Clustering (Hard/Soft).
- It contains set of hyperparameters of the predictor if they exist.
- Express that it can be applied to one data or a collection, i.e for clustering, a distribution or a clustering (hard/soft).
- It HAS to include :
- Single data : ClusterId (Int) / XXXDistribution.
- Collection of data : HardClustering / SoftClustering.
- It HAS to include :
Quality indices
Internals
Externals
Explain concept of ground truth.