more links

_doc/sphinxdoc/source/blog/2018/2018-03-27_new.rst

@@ -0,0 +1,30 @@
+
+.. blogpost::
+    :title: Biblio : articles ajoutés récemment, machine learning
+    :keywords: online training, stochastique gradient descent
+    :date: 2018-03-27
+    :categories: biblio
+
+    Dans :ref:`l-ml2a-ranking` :
+
+    * `Online Learning to Rank with Top-k Feedback <http://www.jmlr.org/papers/volume18/16-285/16-285.pdf>`_
+
+    Les paragraphes :ref:`l-ml2a-sgd`, :ref:`l-basic-gp`,
+    :ref:`l-mlplus-structures` et :ref:`l-ml2a-online-training`
+    ont été révisitées.
+
+    Quelques liens ont été ajoutés dans la section
+    :ref:`l-basic-clustering` notamment
+    l'`Indice de Rand <https://fr.wikipedia.org/wiki/Indice_de_Rand>`_
+    qui sert à comparer la proximité de deux partitions.
+
+    Dans :ref:`l-cluster-graph` :
+
+    L'article
+    `Clustering from General Pairwise Observations with Applications to Time-varying Graphs <http://www.jmlr.org/papers/volume18/15-659/15-659.pdf>`_
+    propose une formulation intéressante du clustering
+    de graphe et explore pour voir où cela mène.
+
+    Un dernier article sur quelques digressions sur la comparaisons
+    de modèles de classification :
+    `Time for a Change: a Tutorial for Comparing Multiple Classifiers Through Bayesian Analysis <http://www.jmlr.org/papers/volume18/16-305/16-305.pdf>`_.

_doc/sphinxdoc/source/ml2a/td2a_mlbasic_clustering.rst

@@ -9,6 +9,8 @@
     :alt: Statistique
     :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
 
+.. _l-basic-clustering:
+
 Clustering
 ++++++++++
 
@@ -22,6 +24,12 @@ Clustering
 * score silhouette
 * clustering de variables catégorielles
 
+*Métriques*
+
+* `Indice de Rand <https://fr.wikipedia.org/wiki/Indice_de_Rand>`_
+* `Silhouette (clustering) <https://en.wikipedia.org/wiki/Silhouette_(clustering)>`_
+* `The Impact of Random Models on Clustering Similarity <http://www.jmlr.org/papers/volume18/17-039/17-039.pdf>`_
+
 *Lectures*
 
 * `A New Algorithm and Theory for Penalized Regression-based Clustering <http://www.jmlr.org/papers/volume17/15-553/15-553.pdf>`_ :
@@ -47,6 +55,7 @@ Clustering
   (implémentation en python `dimitrs/CLTree <https://github.com/dimitrs/CLTree>`_)
 * `Spectral Clustering Based on Local PCA <http://www.jmlr.org/papers/volume18/14-318/14-318.pdf>`_
 * `Brown clustering <https://en.wikipedia.org/wiki/Brown_clustering>`_
+* `Hierarchical Clustering via Spreading Metrics <http://www.jmlr.org/papers/volume18/17-081/17-081.pdf>`_
 
 *Modules*
 

_doc/sphinxdoc/source/ml2a/td2a_mlbasic_machine_learning_cours_de_gael_varoquaux.rst

@@ -61,6 +61,7 @@ ici :
 * `XGBoost: A Scalable Tree Boosting System <http://arxiv.org/pdf/1603.02754v1.pdf>`_, Tianqi Chen, Carlos Guestrin
 * `Round Robin Classification <http://www.jmlr.org/papers/volume2/fuernkranz02a/fuernkranz02a.pdf>`_
 * `ANN-Benchmarks: A Benchmarking Tool for Approximate Nearest Neighbor Algorithms? <http://www.itu.dk/people/maau/additional/sisap2017-preprint.pdf>`_
+* `Time for a Change: a Tutorial for Comparing Multiple Classifiers Through Bayesian Analysis <http://www.jmlr.org/papers/volume18/16-305/16-305.pdf>`_
 
 *Livres*
 

_doc/sphinxdoc/source/ml2a/td2a_mlbasic_ranking.rst

@@ -9,6 +9,8 @@
     :alt: Statistique
     :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
 
+.. _l-ml2a-ranking:
+
 Ranking
 +++++++
 
@@ -21,6 +23,7 @@ Ranking
 * `Data-driven Rank Breaking for Efficient Rank Aggregation <http://www.jmlr.org/papers/volume17/16-209/16-209.pdf>`_
 * `BPR: Bayesian Personalized Ranking from Implicit Feedback <https://arxiv.org/abs/1205.2618>`_
   (applicable également aux systèmes de recommandation)
+* `Online Learning to Rank with Top-k Feedback <http://www.jmlr.org/papers/volume18/16-285/16-285.pdf>`_
 
 *Modules*
 

_doc/sphinxdoc/source/ml2a/td2a_mlplus_autolearning.rst

@@ -36,6 +36,10 @@ Auto-Learning (metalearning)
 * `Random Search for Hyper-Parameter Optimization <http://www.jmlr.org/papers/v13/bergstra12a.html>`_
 * `Reptile: a Scalable Metalearning Algorithm <https://arxiv.org/abs/1803.02999>`_
 
+*POMPD*
+
+* `POMDPs for Dummies <http://www.pomdp.org/tutorial/index.html>`_
+
 *Sites*
 
 * `Machine Learning for Automated Algorithm Design <http://www.ml4aad.org/>`_
@@ -48,3 +52,4 @@ Auto-Learning (metalearning)
 * `TPOT <https://github.com/rhiever/tpot>`_
 * `auto-sklearn <https://github.com/automl/auto-sklearn/>`_
 * `RoBO <https://github.com/automl/RoBO>`_ (bayésien)
+* `POMDPy <https://github.com/pemami4911/POMDPy>`_

_doc/sphinxdoc/source/ml2a/td2a_mlplus_clustering_de_graphes.rst

@@ -4,6 +4,8 @@
     :alt: Statistique
     :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
 
+.. _l-cluster-graph:
+
 Clustering de Graphes
 +++++++++++++++++++++
 
@@ -21,6 +23,7 @@ Clustering de Graphes
 * `Correlation Clustering and Biclustering with Locally Bounded Errors <http://jmlr.org/proceedings/papers/v48/puleo16.pdf>`_
 * `A Unified Framework for Model-based Clustering <http://www.jmlr.org/papers/volume4/zhong03a/zhong03a.pdf>`_
 * `A Tensor Approach to Learning Mixed Membership Community Models <http://jmlr.org/papers/volume15/anandkumar14a/anandkumar14a.pdf>`_
+* `Clustering from General Pairwise Observations with Applications to Time-varying Graphs <http://www.jmlr.org/papers/volume18/15-659/15-659.pdf>`_
 
 *Lectures - métriques*
 

_doc/sphinxdoc/source/ml2a/td2a_mlplus_communautes.rst

@@ -30,3 +30,4 @@ de clustering mais on cherche aussi à en déterminer le centre ou encore
 * `Local Network Community Detection with Continuous Optimization of Conductance and Weighted Kernel K-Means <http://jmlr.org/papers/volume17/16-043/16-043.pdf>`_
 * `Learning Communities in the Presence of Errors <http://www.jmlr.org/proceedings/papers/v49/makarychev16.pdf>`_
 * `Fast Detection of Community Structures using Graph Traversal in Social Networks <https://arxiv.org/pdf/1707.04459.pdf>`_
+* `Community Extraction in Multilayer Networks with Heterogeneous Community Structure <http://www.jmlr.org/papers/volume18/16-645/16-645.pdf>`_

_doc/sphinxdoc/source/ml2a/td2a_mlplus_gaussian_processes.rst

@@ -0,0 +1,29 @@
+
+.. image:: pyeco.png
+    :height: 20
+    :alt: Economie
+    :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-economiste
+
+.. image:: pystat.png
+    :height: 20
+    :alt: Statistique
+    :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
+
+.. _l-basic-gp:
+
+Processus Gaussiens (GP)
+++++++++++++++++++++++++
+
+(à venir)
+
+*Lectures*
+
+* `Introduction aux Processus Gaussiens <http://www.iro.umontreal.ca/~pift6266/A06/cours/030819_talk_lisa_gaussian-process.pdf>`_
+* `Processus gaussiens déformés pour l'apprentissage de zones instationnaires <https://hal.archives-ouvertes.fr/hal-01402259/document>`_
+* `Corresponce entre régression par processus gaussien et splines d'interpolation sous contraintes linéaires de type inégalités, théorie et application <https://tel.archives-ouvertes.fr/tel-01282224/document>`_
+
+*Modules*
+
+* `GPy <https://github.com/SheffieldML/GPy>`_
+* `GPflow <http://gpflow.readthedocs.io/en/latest/index.html>`_
+  (réimplémentation du module précédent sur GPU)

_doc/sphinxdoc/source/ml2a/td2a_mlplus_online_training.rst

@@ -3,11 +3,25 @@
     :alt: Statistique
     :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
 
+.. _l-ml2a-online-training:
+
 Online training
 +++++++++++++++
 
 *(à venir)*
 
 *Lectures*
 
+* `A Survey of Algorithms and Analysis for Adaptive Online Learning <http://www.jmlr.org/papers/volume18/14-428/14-428.pdf>`_
+* `Online Learning and Stochastic Approximations <http://leon.bottou.org/publications/pdf/online-1998.pdf>`_
+* `Sparse Online Learning via Truncated Gradient <http://www.jmlr.org/papers/volume10/langford09a/langford09a.pdf>`_
 * `Fast Rates in Statistical and Online Learning <http://www.jmlr.org/papers/volume16/vanerven15a/vanerven15a.pdf>`_
+* `Mondrian Forests: Efficient Online Random Forests <https://arxiv.org/abs/1406.2673>`_
+* `The Mondrian Process <http://danroy.org/papers/RoyTeh-NIPS-2009.pdf>`_
+* `Extremely randomized trees <http://www.montefiore.ulg.ac.be/~ernst/uploads/news/id63/extremely-randomized-trees.pdf>`_
+* `Approximation Vector Machines for Large-scale Online Learning <http://www.jmlr.org/papers/volume18/16-191/16-191.pdf>`_
+* `Online Learning to Rank with Top-k Feedback <http://www.jmlr.org/papers/volume18/16-285/16-285.pdf>`_
+
+*Modules*
+
+* `scikit-garden <https://github.com/scikit-garden/scikit-garden>`_

_doc/sphinxdoc/source/ml2a/td2a_mlplus_sgd.rst

@@ -0,0 +1,20 @@
+
+.. image:: pystat.png
+    :height: 20
+    :alt: Statistique
+    :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
+
+.. _l-ml2a-sgd:
+
+Stochastique Gradient Descent
++++++++++++++++++++++++++++++
+
+(*à venir*)
+
+*Lectures*
+
+* `Adam: A Method for Stochastic Optimization <https://arxiv.org/abs/1412.6980>`_
+* `HOGWILD!: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent <https://arxiv.org/abs/1106.5730>`_
+* `Sparse Online Learning via Truncated Gradient <http://www.jmlr.org/papers/volume10/langford09a/langford09a.pdf>`_
+* `Stabilized Sparse Online Learning for Sparse Data <http://www.jmlr.org/papers/volume18/16-190/16-190.pdf>`_
+* `On Perturbed Proximal Gradient Algorithms <http://www.jmlr.org/papers/volume18/15-038/15-038.pdf>`_

_doc/sphinxdoc/source/ml2a/td2a_mlplus_structures.rst

@@ -0,0 +1,20 @@
+
+.. image:: pystat.png
+    :height: 20
+    :alt: Statistique
+    :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
+
+.. _l-mlplus-structures:
+
+Apprentissage de structures
++++++++++++++++++++++++++++
+
+(à venir)
+
+*Lectures*
+
+* `Online Bayesian Passive-Aggressive Learning <http://www.jmlr.org/papers/volume18/14-188/14-188.pdf>`_
+
+*Modules*
+
+* `GPy <https://github.com/SheffieldML/GPy>`_

_doc/sphinxdoc/source/ml2a/td2a_mlplus_topic.rst

@@ -0,0 +1,16 @@
+
+.. image:: pystat.png
+    :height: 20
+    :alt: Statistique
+    :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx3/td_2a_notions.html#pour-un-profil-plutot-data-scientist
+
+Topic Modelling
++++++++++++++++
+
+*Lectures*
+
+* `Distributed Algorithms for Topic Models <http://www.jmlr.org/papers/volume10/newman09a/newman09a.pdf>`_
+
+*Modules*
+
+* `Refinery <http://daeilkim.github.io/refinery/>`_

_doc/sphinxdoc/source/ml2a/td2a_mlplus_tres_grande_dimension.rst

@@ -11,3 +11,11 @@ Très grande dimension
 *Lectures*
 
 * `Making Decision Trees Feasible in Ultrahigh Feature and Label Dimensions <http://jmlr.org/papers/volume18/16-466/16-466.pdf>`_
+* `Stabilized Sparse Online Learning for Sparse Data <http://www.jmlr.org/papers/volume18/16-190/16-190.pdf>`_*
+* `Identifying a Minimal Class of Models for High–dimensional Data <http://www.jmlr.org/papers/volume18/16-172/16-172.pdf>`_
+* `Improving Variational Methods via Pairwise Linear Response Identities <http://www.jmlr.org/papers/volume18/16-070/16-070.pdf>`_
+
+*Modules*
+
+* `hypertools <http://hypertools.readthedocs.io/en/latest/>`_
+* `datashader <https://github.com/bokeh/datashader>`_

_doc/sphinxdoc/source/ml2a/td_2a_mlplus.rst

@@ -25,6 +25,8 @@ Machine Learning Avancé
     td2a_mlplus_algorithmes_genetiques_et_autres_variantes
     td2a_mlplus_optimal_transfert
     td2a_mlplus_systeme_de_recommandation_avec_contraintes
+    td2a_mlplus_sgd
+    td2a_mlplus_gaussian_processes
 
 Séries temporelles
 ==================
@@ -105,6 +107,7 @@ les images ou le son.
     :maxdepth: 1
 
     td2a_mlplus_traitement_du_langage
+    td2a_mlplus_topic
     td2a_mlplus_traduction_automatique
     td2a_mlplus_tagging
     td2a_mlplus_ontologie

requirements.txt

@@ -42,7 +42,7 @@ pymmails
 pympler
 pymyinstall>=1.1.1415
 pyqt5
-pyquickhelper>=1.7.2448
+pyquickhelper>=1.7.2508
 pyrsslocal
 python-jenkins
 qgrid