.. image:: pyeco.png
    :height: 20
    :alt: Economie
    :target: http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx/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/helpsphinx/td_2a_notions.html#pour-un-profil-plutot-data-scientist

Transfer Learning
+++++++++++++++++

Le `tranfer learning <https://en.wikipedia.org/wiki/Transfer_learning>`_
consiste à réutiliser un modèle de deep learning déjà appris sur une grande
base de données pour un problème aux dimensions plus modestes.
C'est la première chose à faire quand on commence le deep learning :
c'est souvent très rapide pour des résultats déjà acceptables.
Cela a aussi l'avantage d'être peu coûteux comparé à l'apprentissage
d'un réseau de neurones profond complet sur une grande bases de données.

*Notebooks*

* `Transfer Learning <https://github.com/sdpython/2017_deeplearning_demo/blob/master/Fine_Tuning_Deep_CNNs_with_GPU_rendered.ipynb>`_ (Olivier Grisel)
* `Search images with deep learning <http://www.xavierdupre.fr/app/mlinsights/helpsphinx/notebooks/search_images.html#searchimagesrst>`_

*Lectures - introduction*

* `Building powerful image classification models using very little data <https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html>`_
* `Deep Learning : choisir son framework et entrainer ses modèles dans Azure <http://www.xavierdupre.fr/app/jupytalk/helpsphinx/2017/experience_2017.html>`_,
  présenté avec `Olivier Grisel <http://ogrisel.com/>`_

*Lectures - articles*

* `Unsupervised and Transfer Learning Challenges in Machine Learning, Volume 7 <http://www.mtome.com/Publications/CiML/CiML-v7-book.pdf>`_
* `ICML2011 Unsupervised and Transfer Learning Workshop <http://www.jmlr.org/proceedings/papers/v27/silver12a/silver12a.pdf>`_
* `Transfer Learning <ftp://ftp.cs.wisc.edu/machine-learning/shavlik-group/torrey.handbook09.pdf>`_
* `Deep Learning of Representations for Unsupervised and Transfer Learning <http://www.jmlr.org/proceedings/papers/v27/bengio12a/bengio12a.pdf>`_
* `Unsupervised and Transfer Learning Challenge: a Deep Learning Approach <http://www.jmlr.org/proceedings/papers/v27/mesnil12a/mesnil12a.pdf>`_
* `Transfer Learning by Kernel Meta-Learning <http://www.jmlr.org/proceedings/papers/v27/aiolli12a/aiolli12a.pdf>`_
* `A Survey on Transfer Learning <https://www.cse.ust.hk/~qyang/Docs/2009/tkde_transfer_learning.pdf>`_
* `Domain-Adversarial Training of Neural Networks <http://jmlr.org/papers/volume17/15-239/15-239.pdf>`_
* `Stability and Hypothesis Transfer Learning <http://jmlr.org/proceedings/papers/v28/kuzborskij13.pdf>`_
* `Transfer Learning Decision Forests for Gesture Recognition <http://jmlr.org/papers/volume15/goussies14a/goussies14a.pdf>`_
* `Learning Transferable Features with Deep Adaptation Networks <http://www.jmlr.org/proceedings/papers/v37/long15.pdf>`_
* `Asymmetric Transfer Learning with Deep Gaussian Processes <http://www.jmlr.org/proceedings/papers/v37/kandemir15.pdf>`_
* `Transfer Learning in Sequential Decision Problems: A Hierarchical Bayesian Approach <http://www.jmlr.org/proceedings/papers/v27/wilson12a/wilson12a.pdf>`_
* `Transfer Learning for Reinforcement Learning Domains: A Survey <http://www.jmlr.org/papers/volume10/taylor09a/taylor09a.pdf>`_
* `Unsupervised dimensionality reduction via gradient-based matrix factorization with two adaptive learning rates <http://www.jmlr.org/proceedings/papers/v27/nikulin12a/nikulin12a.pdf>`_
* `Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples <https://arxiv.org/abs/1605.07277>`_
* `Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data <https://arxiv.org/abs/1610.05755>`_
* `Knowledge Concentration: Learning 100K Object Classifiers in a Single CNN <https://arxiv.org/pdf/1711.07607.pdf>`_

*Lectures - hors nn*

* `Transfer Learning Decision Forests for Gesture Recognition <http://jmlr.org/papers/volume15/goussies14a/goussies14a.pdf>`_

*Modèles pré-entraînés*

* `Places CNN <http://places.csail.mit.edu/downloadCNN.html>`_,
  `Pre-release of Places365-CNNs <https://github.com/metalbubble/places365>`_
  (deep learning)
* `CNTK <https://www.microsoft.com/en-us/research/product/cognitive-toolkit/model-gallery/>`_
  (sur `github <https://github.com/Microsoft/CNTK/tree/master/Examples>`_)
* `Model Zoo <https://github.com/BVLC/caffe/wiki/Model-Zoo>`_
* `Model Gallery CNTK <https://www.microsoft.com/en-us/cognitive-toolkit/features/model-gallery/>`_
* `tensorflow/models <https://github.com/tensorflow/models>`_