]> Naucratis ville grecque ou égyptienne, une question de méthode - Art, Archéologie et Antiquité 2018-03-19 2018-03-19T08:37:52Z 2018-03-27 2018-03-27T13:42:54Z python - Running Flask app in a Docker container - Stack Overflow 2018-03-12 2018-03-12T12:55:55Z Présentation rapide de Pandas Deep learning with word embeddings improves biomedical named entity recognition | Bioinformatics | Oxford Academic (2017) 2018-03-05T19:28:35Z 2018-03-05 Open-endedness: The last grand challenge you’ve never heard of - O'Reilly Media 2018-03-30T13:55:20Z 2018-03-30 2018-03-04 2018-03-04T17:15:30Z Extreme Multi-label Learning with Label Features for Warm-start Tagging, Ranking & Recommendation (2018) This paper formulates the extreme classification problem **when predictions need to be made on training points with partially revealed labels**. [SwiftXML pseudo-code](/doc/?uri=https%3A%2F%2Fpdfs.semanticscholar.org%2F873e%2Fea884de581f79b1e783052f8e9fa60726fc8.pdf) **Learns from word2vec features extracted from the tags in addition to the article text features.** IPBES | Science and policy for people and nature 2018-03-26T23:24:38Z 2018-03-26 2018-03-05 2018-03-05T11:29:06Z Principal Component Analysis (PCA) for Feature Selection and some of its Pitfalls · Johannes Otterbach 2018-03-03 2018-03-03T14:27:07Z Machine Learning with Missing Labels: Transductive SVMs 2018-03-06 2018-03-06T14:27:11Z Gaze of time | Timelines and Chronologies curated by the community 2018-03-28T21:54:41Z flask-uploads – Patrick's Software Blog 2018-03-28 Les oiseaux disparaissent des campagnes françaises à une vitesse « vertigineuse » 2018-03-20T08:32:20Z 2018-03-20 Back to the future: Does graph database success hang on query language? | ZDNet 2018-03-13 2018-03-13T14:52:56Z 2015-08-09T06:32:47Z [1508.01991] Bidirectional LSTM-CRF Models for Sequence Tagging 2015-08-09T06:32:47Z In this paper, we propose a variety of Long Short-Term Memory (LSTM) based models for sequence tagging. These models include LSTM networks, bidirectional LSTM (BI-LSTM) networks, LSTM with a Conditional Random Field (CRF) layer (LSTM-CRF) and bidirectional LSTM with a CRF layer (BI-LSTM-CRF). Our work is the first to apply a bidirectional LSTM CRF (denoted as BI-LSTM-CRF) model to NLP benchmark sequence tagging data sets. We show that the BI-LSTM-CRF model can efficiently use both past and future input features thanks to a bidirectional LSTM component. It can also use sentence level tag information thanks to a CRF layer. The BI-LSTM-CRF model can produce state of the art (or close to) accuracy on POS, chunking and NER data sets. In addition, it is robust and has less dependence on word embedding as compared to previous observations. Wei Xu Bidirectional LSTM-CRF Models for Sequence Tagging Zhiheng Huang 2018-03-05T19:03:20Z 2018-03-05 1508.01991 Zhiheng Huang Kai Yu 2018-03-18 How Trump Consultants Exploited the Facebook Data of Millions - The New York Times 2018-03-18T10:26:39Z Codes of Interest: Using Bottleneck Features for Multi-Class Classification in Keras and TensorFlow 2018-03-04T16:49:06Z 2018-03-04 2018-03-24 2018-03-24T14:28:29Z The Best Alternative For Every Facebook Feature | WIRED library/python - Docker Hub 2018-03-26 2018-03-26T08:33:35Z 2018-03-12T11:22:58Z GitHub - anvaka/word2vec-graph: Exploring word2vec embeddings as a graph of nearest neighbors 2018-03-12 Semantic hashing using tags and topic modeling (2013) 2018-03-22 2018-03-22T00:41:03Z Semantic Hashing using Tags and Topic Modeling, to incorporate both the tag information and the similarity information from probabilistic topic modeling. [Comments about the paper](https://sutheeblog.wordpress.com/2016/10/28/paper-reading-semantic-hashing-using-tags-and-topic-modeling-sigir13/). [Code on Github](https://github.com/zhuoxiongzhao/code-for-SHTTM) 2018-03-05T18:34:13Z Notes from Coursera Deep Learning courses by Andrew Ng 2018-03-05 Lorsque Maillart, juge d’enfer, menait A Montfaucon Semblançay l’âme rendre, A votre avis, lequel des deux tenait Meilleur maintien ? Pour le vous faire entendre, Maillart sembla l’homme que mort va prendre, Et Semblançay fut si ferme vieillard Que l’on cuidait pour vray qu’il menast pendre A Montfaucon le lieutenant Maillart. (Clément Marot) 2018-03-26 2018-03-26T09:34:09Z La condamnation de Semblançay : une erreur judiciaire ? | autourdemesromans.com 2018-03-20 Maximilian Lam Word2Bits - Quantized Word Vectors 2018-03-15T09:21:34Z 2018-03-31T08:45:59Z Maximilian Lam We show that high quality quantized word vectors using 1-2 bits per parameter can be learned by introducing a quantization function into Word2Vec. We furthermore show that training with the quantization function acts as a regularizer 1803.05651 [1803.05651] Word2Bits - Quantized Word Vectors 2018-03-20T17:36:21Z Word vectors require significant amounts of memory and storage, posing issues to resource limited devices like mobile phones and GPUs. We show that high quality quantized word vectors using 1-2 bits per parameter can be learned by introducing a quantization function into Word2Vec. We furthermore show that training with the quantization function acts as a regularizer. We train word vectors on English Wikipedia (2017) and evaluate them on standard word similarity and analogy tasks and on question answering (SQuAD). Our quantized word vectors not only take 8-16x less space than full precision (32 bit) word vectors but also outperform them on word similarity tasks and question answering. 2018-03-16 2018-03-16T16:45:54Z In Depth: Gaussian Mixture Models | Python Data Science Handbook The non-probabilistic nature of [k-means](/tag/k_means_clustering) and its use of simple distance-from-cluster-center to assign cluster membership leads to poor performance for many real-world situations. We take a look at Gaussian mixture models (GMMs), which can be viewed as an extension of the ideas behind k-means, but can also be a powerful tool for estimation beyond simple clustering. A Gaussian mixture model (GMM) attempts to find a mixture of multi-dimensional Gaussian probability distributions that best model any input dataset. In the simplest case, GMMs can be used for finding clusters in the same manner as k-means. But because GMM contains a probabilistic model under the hood, it is also possible to find probabilistic cluster assignments (in Scikit-Learn, using predict_proba) > Though GMM is often categorized as a clustering algorithm, fundamentally it is an algorithm for density estimation. That is to say, the result of a GMM fit to some data is technically not a clustering model, but a generative probabilistic model describing the distribution of the data. -> a natural means of determining the optimal number of components for a given dataset The Building Blocks of Interpretability 2018-03-07 2018-03-07T14:32:55Z > Interpretability techniques are normally studied in isolation. We explore the powerful interfaces that arise when you combine them — and the rich structure of this combinatorial space. 2018-03-29T00:37:46Z 2018-03-29 Intelligence artificielle : ce qu’il faut retenir du rapport de Cédric Villani 2018-03-28T22:02:15Z Sommet intelligence artificielle à Paris 2018-03-28 2018-03-04 Banksy documentary: Welcome to the Banksy art hotel in Bethlehem - YouTube 2018-03-04T10:09:04Z 2018-03-15 2018-03-15T13:26:54Z Big data for the people: it's time to take it back from our tech overlords | Technology | The Guardian 2018-03-04 Hsiang-Fu Yu [1307.5101] Large-scale Multi-label Learning with Missing Labels Hsiang-Fu Yu The multi-label classification problem has generated significant interest in recent years. However, existing approaches do not adequately address two key challenges: (a) the ability to tackle problems with a large number (say millions) of labels, and (b) the ability to handle data with missing labels. In this paper, we directly address both these problems by studying the multi-label problem in a generic empirical risk minimization (ERM) framework. Our framework, despite being simple, is surprisingly able to encompass several recent label-compression based methods which can be derived as special cases of our method. To optimize the ERM problem, we develop techniques that exploit the structure of specific loss functions - such as the squared loss function - to offer efficient algorithms. We further show that our learning framework admits formal excess risk bounds even in the presence of missing labels. Our risk bounds are tight and demonstrate better generalization performance for low-rank promoting trace-norm regularization when compared to (rank insensitive) Frobenius norm regularization. Finally, we present extensive empirical results on a variety of benchmark datasets and show that our methods perform significantly better than existing label compression based methods and can scale up to very large datasets such as the Wikipedia dataset. 2018-03-04T17:05:39Z Prateek Jain Large-scale Multi-label Learning with Missing Labels 2013-11-25T16:57:43Z 2013-07-18T23:55:55Z Inderjit S. Dhillon 1307.5101 Purushottam Kar Biodiversité : l’urgence du politique 2018-03-26 2018-03-26T23:17:12Z 2016-04-22T16:34:30Z 2016-04-22T16:34:30Z 2018-03-03 1604.06737 Cheng Guo Felix Berkhahn Entity Embeddings of Categorical Variables 2018-03-03T17:13:44Z Cheng Guo [1604.06737] Entity Embeddings of Categorical Variables > We map categorical variables in a function approximation problem into Euclidean spaces, which are the entity embeddings of the categorical variables. The mapping is learned by a neural network during the standard supervised training process. Entity embedding not only reduces memory usage and speeds up neural networks compared with one-hot encoding, but more importantly by mapping similar values close to each other in the embedding space it reveals the intrinsic properties of the categorical variables We map categorical variables in a function approximation problem into Euclidean spaces, which are the entity embeddings of the categorical variables. The mapping is learned by a neural network during the standard supervised training process. Entity embedding not only reduces memory usage and speeds up neural networks compared with one-hot encoding, but more importantly by mapping similar values close to each other in the embedding space it reveals the intrinsic properties of the categorical variables. We applied it successfully in a recent Kaggle competition and were able to reach the third position with relative simple features. We further demonstrate in this paper that entity embedding helps the neural network to generalize better when the data is sparse and statistics is unknown. Thus it is especially useful for datasets with lots of high cardinality features, where other methods tend to overfit. We also demonstrate that the embeddings obtained from the trained neural network boost the performance of all tested machine learning methods considerably when used as the input features instead. As entity embedding defines a distance measure for categorical variables it can be used for visualizing categorical data and for data clustering. « Hey Cricket Australia, show some balls » 2018-03-27 2018-03-27T12:05:40Z Le scandale qui secoue le cricket australien 2018-03-31T10:02:36Z En Syrie, le plus ancien palais de l’humanité détruit par l’organisation Etat islamique 2018-03-31 2018-03-21 2018-03-21T09:32:06Z Facebook/Cambridge Analytica: Privacy lessons and a way forward | Internet Policy Research Initiative @ MIT Intro to text classification with Keras: automatically tagging Stack Overflow posts | Google Cloud Big Data and Machine Learning Blog 2018-03-04 2018-03-04T16:59:49Z 2018-03-12T12:51:52Z 2018-03-12 Google's Machine Learning Crash Course  |  Google Developers 2018-03-12 2018-03-12T11:13:44Z GitHub - spotify/annoy: Approximate Nearest Neighbors in C++/Python optimized for memory usage and loading/saving to disk [Enquête] Soja: quand la déforestation s'invite dans nos assiettes - Amériques - RFI 2018-03-26T17:46:05Z 2018-03-26 Aucune mention de l'origine du soja qui alimente les animaux d'élevage français n'est obligatoire sur les étiquettes. 2018-03-26T08:26:50Z 2018-03-26 How to Build and Deploy a Python Application on Docker | Distelli 2018-03-27 2018-03-27T09:39:00Z GitHub - solid/solid: Solid - Re-decentralizing the web (project directory) Revealed: 50 million Facebook profiles harvested for Cambridge Analytica in major data breach | News | The Guardian 2018-03-18 2018-03-18T10:38:51Z 2018-03-09T13:54:21Z 2018-03-09 Flask (A Python Microframework) 2018-03-28 2018-03-28T23:57:41Z AI for humanity Chris Dyer 2018-03-05T18:40:55Z Neural architectures for NER that use no language-specific resources or features beyond a small amount of supervised training data and unlabeled corpora. > Our models rely on two sources of information about words: character-based word representations learned from the supervised corpus and unsupervised word representations learned from unannotated corpora Sandeep Subramanian 2018-03-05 Kazuya Kawakami [1603.01360] Neural Architectures for Named Entity Recognition Miguel Ballesteros 2016-03-04T06:36:29Z State-of-the-art named entity recognition systems rely heavily on hand-crafted features and domain-specific knowledge in order to learn effectively from the small, supervised training corpora that are available. In this paper, we introduce two new neural architectures---one based on bidirectional LSTMs and conditional random fields, and the other that constructs and labels segments using a transition-based approach inspired by shift-reduce parsers. Our models rely on two sources of information about words: character-based word representations learned from the supervised corpus and unsupervised word representations learned from unannotated corpora. Our models obtain state-of-the-art performance in NER in four languages without resorting to any language-specific knowledge or resources such as gazetteers. Guillaume Lample Neural Architectures for Named Entity Recognition 1603.01360 Guillaume Lample 2016-04-07T15:09:36Z ML-knn: A lazy learning approach to multi-label learning (2007) 2018-03-18 2018-03-18T10:54:01Z a lazy learning approach derived from the traditional [k-nearest neighbors algorithm](tag:k_nearest_neighbors_algorithm) > for each unseen instance, its K nearest neighbors in the training set are firstly identified. After that, based on statistical information gained from the label sets of these neighboring instances, i.e. the number of neighboring instances belonging to each possible class, [Maximum a posteriori (MAP)](tag:maximum_a_posteriori_estimation) principle is utilized to determine the label set for the unseen instance. Implemented in [scikit-multilearn](http://scikit.ml/api/skmultilearn.adapt.mlknn.html), in [java](https://github.com/lefman/mulan-extended/blob/master/mulan/src/mulan/classifier/lazy/MLkNN.java) > the first lazy approach proposed specifically for multi-label classification. This is also a binary relevance approach which considers each label independently as a binary classification problem. Instead of a standard k-NN method, however, MLkNN uses the maximum a-posteriori (MAP) (Kelleher et al., 2015) approach combined with k-NN. [src](https://pdfs.semanticscholar.org/af9b/33da37d290c063cd826ab5923d96892a9767.pdf) 2018-03-22 2018-03-22T00:13:07Z LSH Forest: Self-Tuning Indexes for Similarity Search (2005) 2018-03-12 2018-03-12T13:16:46Z L’ex-Premier ministre du Japon, Naoto Kan, raconte la catastrophe de Fukushima MA Thesis 2018-03-05 2018-03-05T11:41:06Z Examination of machine learning methods for multi-label classification of intellectual property documents (2017) NLP: Requests for Research 2018-03-04T16:38:14Z 2018-03-04 A survey of named entity recognition and classification (2006) 2018-03-05T01:35:29Z 2018-03-05 > we propose a word representation that includes both the word-level and character-level information 2018-03-06 2018-03-06T11:08:23Z Effective Word Representation for Named Entity Recognition (2017) 2018-03-04 2018-03-04T12:05:41Z Aux origines de CRISPR 2018-03-10 REST is the new SOAP – freeCodeCamp 2018-03-10T09:23:04Z [a response](https://philsturgeon.uk/api/2017/12/18/rest-confusion-explained/) 2018-03-06T11:59:39Z SPACY'S ENTITY RECOGNITION MODEL: incremental parsing with Bloom embeddings & residual CNNs - YouTube 2018-03-06 2018-03-26 2018-03-26T23:22:37Z Worsening Worldwide Land Degradation Now ‘Critical’, Undermining Well-Being of 3.2 Billion People | IPBES 2018-03-26 Beginner’s guide to Data Science — Python + Docker – Towards Data Science 2018-03-26T08:29:44Z 2018-03-31T19:49:02Z Emmanuel Macron Q&A: France's President Discusses Artificial Intelligence Strategy | WIRED 2018-03-31 How to manipulate Facebook and Twitter instead of letting them manipulate you - MIT Technology Review 2018-03-21T09:47:45Z 2018-03-21 2018-03-16T23:57:54Z Label Embedding Trees for Large Multi-Class Tasks (2010) 2018-03-16 > Multi-class classification becomes challenging at test time when the number of classes is very large and testing against every possible class can become computationally infeasible. **This problem can be alleviated by imposing (or learning) a structure over the set of classes**. We propose **an algorithm for learning a tree-structure of classifiers** which, by optimizing the overall tree loss, provides superior accuracy to existing tree labeling methods. We also propose **a method that learns to embed labels in a low dimensional space** that is faster than non-embedding approaches and has superior accuracy to existing embedding approaches. Finally we combine the two ideas resulting in the label embedding tree that outperforms alternative methods including One-vs-Rest while being orders of magnitude faster. 2017-07-03T06:37:01Z 2018-03-16 Uses [Deep Canonical Correlation Analysis](/tag/deep_canonical_correlation_analysis) and autoencoder structures to **learn a latent subspace from both feature and label domains** for multi-label classification. (several implementations on github) Chih-Kuan Yeh Wei-Jen Ko 2017-07-03T06:37:01Z Chih-Kuan Yeh 2018-03-16T23:37:58Z 1707.00418 Multi-label classification is a practical yet challenging task in machine learning related fields, since it requires the prediction of more than one label category for each input instance. We propose a novel deep neural networks (DNN) based model, Canonical Correlated AutoEncoder (C2AE), for solving this task. Aiming at better relating feature and label domain data for improved classification, we uniquely perform joint feature and label embedding by deriving a deep latent space, followed by the introduction of label-correlation sensitive loss function for recovering the predicted label outputs. Our C2AE is achieved by integrating the DNN architectures of canonical correlation analysis and autoencoder, which allows end-to-end learning and prediction with the ability to exploit label dependency. Moreover, our C2AE can be easily extended to address the learning problem with missing labels. Our experiments on multiple datasets with different scales confirm the effectiveness and robustness of our proposed method, which is shown to perform favorably against state-of-the-art methods for multi-label classification. Wei-Chieh Wu [1707.00418] Learning Deep Latent Spaces for Multi-Label Classification Learning Deep Latent Spaces for Multi-Label Classification Yu-Chiang Frank Wang 2018-03-29 2018-03-29T16:45:36Z GitHub - ijkilchenko/Fuzbal: Chrome extension: Gives Ctrl+F like find results which include non-exact (fuzzy) matches using string edit-distance and GloVe/Word2Vec. Also searches by regular expressions. Learning to write programs that generate images | DeepMind 2018-03-28T12:11:42Z 2018-03-28 This ability to interpret objects through the tools that created them gives us a richer understanding of the world and is an important aspect of our intelligence. Greenpeace alerte sur le boom de la pêche au krill en Antarctique 2018-03-13T14:11:33Z 2018-03-13 What worries me about AI – François Chollet – Medium 2018-03-29T19:38:10Z 2018-03-29 what really worries me when it comes to AI: the highly effective, highly scalable manipulation of human behavior that AI enables, and its malicious use by corporations and governments 2018-03-27 2018-03-27T14:06:06Z Structuring Your Project — The Hitchhiker's Guide to Python 2018-03-15 2018-03-15T13:55:54Z Speech and Language Processing Climate change is a disaster foretold, just like the first world war | Jeff Sparrow | Opinion | The Guardian 2018-03-12 2018-03-12T13:03:43Z Utiliser un dictionnaire chinois - 东游记 2018-03-04 2018-03-04T11:03:02Z 2018-03-05T18:51:35Z 2018-03-05 bi-LSTM + CRF with character embeddings for NER and POS. [linked from here](http://nlp.town/blog/ner-and-the-road-to-deep-learning/) Sequence Tagging with Tensorflow Multiclass and multilabel algorithms — scikit-learn documentation 2018-03-17T14:38:19Z 2018-03-17