2b:null 2c:["$","title",null,{"children":"Autonomous Entailment | James Anderson | Observable"}] 31:T99f,mdfn `The problem is to derive specific values for some relationship when the only stored data pertains to relationships which should be similar but are not identical. In the context of SPARQL processing, this is known as [entailment](https://www.w3.org/TR/sparql11-entailment/). A simple SPARQL evaluation involves interpreting the triples in both the RDF data and the query as designators for node graphs and matching the query graph against the data graph to yield bindings for any variables. Entailment adds meaning to the data by incorporating into the interpretation relationships which the processor derives from rules or infers from data statistics. Rule-based entailment involves definitions independent of the data. These may be fixed collections, such as [RDFS-](https://www.w3.org/TR/sparql11-entailment/#RDFSEntRegime) or [D-entailment](https://www.w3.org/TR/sparql11-entailment/#DEntRegime), or them may involve a meta-language, such as [OWL](https://www.w3.org/TR/sparql11-entailment/#OWLRDFBSEntRegime) which permits to define the rules. Autonomous entailment is based on rules extracted from the data itself, based on "a joint probability distribution of input (measurements) and output (say class labels) variables."[^shashua2008] Curated rules are managed manually in an independent repository and incorporated by reference. Autonomous rules are computed from the repository's content as a side-effect of operations on it, whereby the respective rule model and any hyperparameters which govern the model extraction are also stored in an independent repository. The models are computed "automatically" through application of a selected learning process while the hyperparameters are managed manually. The learning process involves queries which embed values from the base repository model into the learning process' feature value space and analogous queries which project back from that space into the repository model. These queries are associated with the learned model repository. Taken together, these elements constitute the PMD (preprocessing, modeling, deployment) phases of a "machine learning pipeline". - data repository - preprocessing query - embedding query : maps of a discrete — categorical — variable to a vector of continuous numbers - model repository [^koehrsen2018] - extraction query - hyperparameters - model parameters - vocabulary repository : data which informs the embedding and extraction processes `32:T4a3,## Alternative Implementations >

library	classification	regression	support vector machines	decision trees	ensemble learning	neural networks
scicat-learn	SGDClassifier OneVsOneClassifier OneVsRestClassifier KNeighborsClassifier	LinearRegression SGDRegressor LogisticRegression PolynomialFeatures Ridge Lasso ElasticNet CatBoostRegressor GaussianNB KernelRidge ElasticNet BayesianRidge	SVC LinearSVC SVR	DecisionTreeClassifier	RandomForestClassifier VotingClassifier GradientBoostingClassifier LGBMClassifier CatBoostClassifier XGBClassifier LGBMRegressor CatBoostRegressor XGBRegressor
clml
mgl: classification and neural networks from gabor melis						programmatic definition and controlflow

33:[{"id":"6e2f745c7da8a3911f0c345f3744068cec6251fae502639c17b21f496f8b6774e2293c46707b1de8a11fe34e2bdf6092bf66cedb9c42a7dca19e557e87331e81","href":"https://static.observableusercontent.com/files/6e2f745c7da8a3911f0c345f3744068cec6251fae502639c17b21f496f8b6774e2293c46707b1de8a11fe34e2bdf6092bf66cedb9c42a7dca19e557e87331e81","name":"algo-class-hierarchy.circos+svg.svg","type":"image/svg+xml","lastModified":1636369525040.726,"size":2316905},{"id":"eb06ea08e372bbb0c8120fb6e39d43c1c07501e33d198c9ad4d238e24377a51c9baf1f7e8a3211381c18d940a642bd9a612cb55c6bec735f17d54cd858cc75c6","href":"https://static.observableusercontent.com/files/eb06ea08e372bbb0c8120fb6e39d43c1c07501e33d198c9ad4d238e24377a51c9baf1f7e8a3211381c18d940a642bd9a612cb55c6bec735f17d54cd858cc75c6","name":"algo-class-hierarchy.dot+svg.svg","type":"image/svg+xml","lastModified":1636369479082.647,"size":287414},{"id":"c90dfe40187b4a000e66ad248ebec0554f02c993b08454fa11b518ed22ec06367e904939b403ffe23a04b4fdd7a809a9f42a6433a7a54573179a8f6ac62a4fcc","href":"https://static.observableusercontent.com/files/c90dfe40187b4a000e66ad248ebec0554f02c993b08454fa11b518ed22ec06367e904939b403ffe23a04b4fdd7a809a9f42a6433a7a54573179a8f6ac62a4fcc","name":"feature-store.pdf","type":"application/pdf","lastModified":1672626901643.1,"size":21806},{"id":"40609983f0ca040322f8bf40ff25e3a2aa9c070fff61492ad478b54b73a1400fc9620c2100b0dc4a90b4694391ac9073b362d78254678c7c26a9fa3368d37e9b","href":"https://static.observableusercontent.com/files/40609983f0ca040322f8bf40ff25e3a2aa9c070fff61492ad478b54b73a1400fc9620c2100b0dc4a90b4694391ac9073b362d78254678c7c26a9fa3368d37e9b","name":"feature-store.svg","type":"image/svg+xml","lastModified":1672627002510.019,"size":13251}] 2f:{"id":"29b660619b3d2263","slug":"autonomous-entailment","title":"Autonomous Entailment","thumbnail":"ce4f6804df8f12c36f3f2aaae9f189ce7aea59d55b67361e2eaa2c39363972fc","access_level":"public","create_time":"$D2021-11-03T15:54:03.036Z","update_time":"$D2024-08-13T00:17:36.240Z","view_time":null,"first_public_time":"$D2023-01-01T14:18:44.325Z","fork":false,"stars":0,"starred":false,"owner":{"type":"team","id":"b452fd69977f5701","login":"lomoramic","name":"James Anderson","avatar_url":"https://avatars-next.observableusercontent.com/dc520bbc733259f2913522cf1d50607777b98a3083f290a6f182534561896497","tier":"starter_2024"},"version":168,"publish_version":null,"role":null,"authored":false,"authors":[{"type":"individual","id":"987c90ad70c449c1","login":"lomoramic","name":"James Anderson","avatar_url":"https://avatars-next.observableusercontent.com/dc520bbc733259f2913522cf1d50607777b98a3083f290a6f182534561896497","tier":null}],"members":[{"type":"individual","id":"987c90ad70c449c1","login":"lomoramic","name":"James Anderson","avatar_url":"https://avatars-next.observableusercontent.com/dc520bbc733259f2913522cf1d50607777b98a3083f290a6f182534561896497","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"subscribed":null,"forks":0,"collection":null,"type":"notebook","body":{"title":"Autonomous Entailment","cells":[{"id":0,"value":"# Autonomous Entailment","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined"},{"id":3,"value":"$31","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":58,"value":"## Learning Process\n- post to preprocessing query\n- generate embedding\n - for initial window\n - for new data model / incremental snapshot\n- refine prediction model","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined"},{"id":55,"value":"## Query Process\n- embed the query data model terms to locations\n- compute predictions in embedded space\n- apply model\n - determine nearest neighbours\n - predict result variables\n- extract the locations to data model terms","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined"},{"id":68,"value":"mdfn `\n## Management\n\nThe learning and query process must track and externalize\n- correlation between data repository version and model repository version\n- data model statistics\n- entailment model quality statistics\n- [cd4ml](https://martinfowler.com/articles/cd4ml.html):\n - validate data : internal learning data integrity, online data equivalence\n - component integration : result v/s initial dimension correspondence\n - model quality : accuracy, precision, recall\n - model bias : relative statistics of learning and on-line data\n`","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":157,"value":"## Process Flow\n\nA model is generated from a dataset by defining the learing steps (embedding, meta-parameters) and initiating the learning process.\n\nThe result is available in the form of a location, which is incorporated into queries as a functional predicate.\n${await(featureStore.image())}","pinned":true,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined"},{"id":160,"value":"featureStore = FileAttachment(\"feature-store.svg\")","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":97,"value":"mdfn `\n## Ontology for machine Learning\n\nViews of the machine learning ontology from \"A Machine Learning Ontology\"[^braga2020], as a network graph and as a circos diagram.\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":99,"value":"FileAttachment(\"algo-class-hierarchy.dot+svg.svg\").image()","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":105,"value":"FileAttachment(\"algo-class-hierarchy.circos+svg.svg\").image()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":72,"value":"$32","pinned":true,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined"},{"id":25,"value":"viewof footnotes = footlist([\n {ref: 'shashua2008', text: `Introduction to Machine Learning, p1`},\n {ref: 'koehrsen2018', text: `Koehrsen [Neural Network Embeddings Explained](https://towardsdatascience.com/neural-network-embeddings-explained-4d028e6f0526)`},\n {ref: 'braga2020', text: `Braga, J., Dias, J.L. and Regateiro, F., 2020. [A MACHINE LEARNING ONTOLOGY](https://frenxiv.org/rc954/download/?format=pdf)`}\n ])","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":20,"value":"----","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined"},{"id":22,"value":"import {mdfn, footlist, footnoteCSS} with { footnotes as footnotes } from '@formsandlines/linked-markdown-style-footnotes-with-tooltips'\n","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":148,"value":"footnoteCSS","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"}],"files":"$W33","stdlib":"1","titling":"auto"}} 30:"$2f:body"