]> 2017-12-17T11:55:50Z The Wikipedia Competitor That's Harnessing Blockchain For Epistemological Supremacy | WIRED 2017-12-17 2017-12-10 2017-12-10T14:58:36Z Seigneurie de Bellême 2017-12-16 2017-12-16T15:03:15Z When Debate Is Futile: Bertrand Russell’s Remarkable Response to a Fascist’s Provocation – Brain Pickings 2017-12-04 Optimization for Deep Learning Highlights in 2017 2017-12-04T12:11:44Z 2017-12-06 Taming Recurrent Neural Networks for Better Summarization | Abigail See 2017-12-06T23:32:43Z Les enjeux de mise en œuvre opérationnelle de l’intelligence artificielle dans les grandes entreprises 2017-12-18 2017-12-18T10:47:41Z > il ne faut pas oublier qu’il y a toujours deux dimensions dans le sujet de l’IA : > >- Celle de l’homme augmenté avec l’analytics, l’aide à la décision, le machine learning >- Celle de l’homme remplacé avec des tâches entières de l’entreprise qui sont déléguées à la décision de robots intelligents. > > Nous ne sommes pas encore à la phase de remplacement, mais plutôt à celle d’augmentation. A propos de chatbots dans les RH : > limiter au maximum l’augmentation de l’homme pour ne pas avoir à baisser les effectifs. ! The Merge - Sam Altman 2017-12-07T18:31:41Z 2017-12-07 Learning Entity Embeddings in one breath – Apil Tamang – Medium 2017-12-03 2017-12-03T10:52:02Z example of a recommender system, with person-item matrix It’s what corporations were designed to do: Let a bunch of people get together, take some strategic risks they might otherwise not take, and then make sure none of them is devastated individually if things go south. Why Aren’t Any Bankers in Prison for Causing the Financial Crisis? - The Atlantic 2017-12-17T00:37:56Z 2017-12-17 10 Breakthrough Technologies 2017 - MIT Technology Review 2017-12-31 2017-12-31T10:43:46Z 2017-12-08 2017-12-08T17:16:00Z Kassav ' - Biographie, discographie et fiche artiste – RFI Musique A Latent Variable Model Approach to PMI-basedWord Embeddings 2017-12-07T16:20:59Z 2017-12-07 (improves on [Yoav Goldberg](/tag/yoav_goldberg)'s findings) 2017-12-16 Precious Gems Bear Messages From Earth’s Molten Heart - The New York Times 2017-12-16T18:06:53Z Énergie : les promesses de l'hydrogène | CNRS Le journal 2017-12-08T23:32:32Z 2017-12-08 Using Artificial Intelligence to Augment Human Intelligence 2017-12-05T18:28:06Z 2017-12-05 By creating user interfaces which let us work with the representations inside machine learning models, we can give people new tools for reasoning. 2017-12-01 Artificial intelligence goes bilingual—without a dictionary | Science | AAAS 2017-12-01T01:25:34Z 2017-12-13 2017-12-13T11:22:56Z Deep Learning for NLP, advancements and trends in 2017 - Tryolabs Blog « Paradise Papers » : faire la morale ne suffit pas 2017-12-05T19:24:55Z 2017-12-05 2017-12-13 2017-12-13T11:11:04Z Everything is a Model | Delip Rao Comments on the [“The Case for Learned Index Structures”](https://arxiv.org/abs/1712.01208v1) paper 2017-12-11T19:25:09Z Tim Kraska Tim Kraska 2018-04-30T07:54:41Z > we believe that the idea of replacing core components of a data management system through learned models has far reaching implications for future systems designs > > Indexes are models: a B-Tree-Index can be seen as a model to map a key to the position of a record within a sorted array, a Hash-Index as a model to map a key to a position of a record within an unsorted array, and a BitMap-Index as a model to indicate if a data record exists or not. In this exploratory research paper, we start from this premise and posit that all existing index structures can be replaced with other types of models, including deep-learning models, which we term learned indexes. 2017-12-04T17:18:41Z The Case for Learned Index Structures Jeffrey Dean Alex Beutel Neoklis Polyzotis 2017-12-11 1712.01208 Indexes are models: a B-Tree-Index can be seen as a model to map a key to the position of a record within a sorted array, a Hash-Index as a model to map a key to a position of a record within an unsorted array, and a BitMap-Index as a model to indicate if a data record exists or not. In this exploratory research paper, we start from this premise and posit that all existing index structures can be replaced with other types of models, including deep-learning models, which we term learned indexes. The key idea is that a model can learn the sort order or structure of lookup keys and use this signal to effectively predict the position or existence of records. We theoretically analyze under which conditions learned indexes outperform traditional index structures and describe the main challenges in designing learned index structures. Our initial results show, that by using neural nets we are able to outperform cache-optimized B-Trees by up to 70% in speed while saving an order-of-magnitude in memory over several real-world data sets. More importantly though, we believe that the idea of replacing core components of a data management system through learned models has far reaching implications for future systems designs and that this work just provides a glimpse of what might be possible. [1712.01208] The Case for Learned Index Structures Ed H. Chi 2017-12-30T11:07:53Z 2017-12-30 A new idea called the “information bottleneck” is helping to explain the puzzling success of today’s artificial-intelligence algorithms — and might also explain how human brains learn. New Theory Cracks Open the Black Box of Deep Learning | Quanta Magazine 2017-12-10 2017-12-10T19:54:07Z What They Don't Tell You About Data Science. 1: You Are a Software Engineer First - DS lore 2017-12-18T10:50:06Z 2017-12-18 Etude CIGREF : enjeux de mise en œuvre de l’Intelligence Artificielle pour l’entreprise… – CIGREF 2017-12-04T09:03:51Z Solving the Neural Code Conundrum: Digital or Analog? - MIT Technology Review They analyse a neuronal signal and then try to reproduce it using the empirical Bayes model and then using the hidden Markov model. They then decide whether it is digital or analog depending on the model that best simulates the characteristics of the original signal. 2017-12-04 The Son (Nesbø novel) 2017-12-29T19:05:26Z 2017-12-29 6,909 living languages on Earth. 414 of those account for 94% of humanity 2017-12-08 2017-12-08T14:28:40Z Number of languages Paradigm shifts for the decentralized Web | Ruben Verborgh 2017-12-21T00:42:04Z 2017-12-21 [1711.07128] Hello Edge: Keyword Spotting on Microcontrollers 2017-11-20T03:19:03Z 2017-12-15 Naveen Suda 2018-02-14T19:24:55Z Keyword spotting (KWS) is a critical component for enabling speech based user interactions on smart devices. It requires real-time response and high accuracy for good user experience. Recently, neural networks have become an attractive choice for KWS architecture because of their superior accuracy compared to traditional speech processing algorithms. Due to its always-on nature, KWS application has highly constrained power budget and typically runs on tiny microcontrollers with limited memory and compute capability. The design of neural network architecture for KWS must consider these constraints. In this work, we perform neural network architecture evaluation and exploration for running KWS on resource-constrained microcontrollers. We train various neural network architectures for keyword spotting published in literature to compare their accuracy and memory/compute requirements. We show that it is possible to optimize these neural network architectures to fit within the memory and compute constraints of microcontrollers without sacrificing accuracy. We further explore the depthwise separable convolutional neural network (DS-CNN) and compare it against other neural network architectures. DS-CNN achieves an accuracy of 95.4%, which is ~10% higher than the DNN model with similar number of parameters. 2017-12-15T09:04:47Z Vikas Chandra Yundong Zhang Hello Edge: Keyword Spotting on Microcontrollers Yundong Zhang Liangzhen Lai 1711.07128 2017-12-30 DSSM ("Deep Semantic Similarity Model") - Microsoft Research 2017-12-30T02:04:48Z Deep neural network modeling technique for representing text strings (sentences, queries, predicates, entity mentions, etc.) in a continuous semantic space and modeling semantic similarity between two text strings 2017-12-19 > a suite of five papers that support the emerging realization that neuroevolution, where neural networks are optimized through evolutionary algorithms, is also an effective method to train deep neural networks for reinforcement learning (RL) problems. Welcoming the Era of Deep Neuroevolution - Uber Engineering Blog 2017-12-19T09:26:01Z UK cannot have a special deal for the City, says EU's Brexit negotiator | Politics | The Guardian In a blow to remain campaigners, Barnier contends that the UK would be unable to revoke article 50 unilaterally 2017-12-19 2017-12-19T14:06:45Z 2017-12-12 2017-12-12T21:46:59Z The Paradox of the Proof | Project Wordsworth 2017-12-14T21:32:21Z 2017-12-14 Gouvernance de l'intelligence artificielle dans les grandes entreprises Sécurité nucléaire : le grand mensonge | ARTE Cinema 2017-12-07 2017-12-07T00:10:57Z <https://www.dailymotion.com/video/x6f2z3t> **the norm of a word vector is somewhat related to the overall frequency** of which words occur in the training corpus (so a common word like "frog" will still be similar to a less frequent word like "Anura" which is it's scientific name) (Hence the use of cosine-distance) > That the inner product relates to the PMI between the vectors is for the most part an empirical result and there is very little theoretical background behind this finding (fastText) Euclidean distance instead of cosine-similarity? 2017-12-07T16:06:35Z 2017-12-07 2017-12-03 2017-12-03T14:18:06Z Naked Blood 2017-12-22 2017-12-22T10:09:41Z facebookresearch/MUSE: A library for Multilingual Unsupervised or Supervised word Embeddings Les saboteurs de la centrale de Doel toujours à l’intérieur ? - Greenpeace France 2017-12-07T00:39:19Z 2017-12-07 2017-12-30T02:10:49Z Learning Deep Structured Semantic Models for Web Search using Clickthrough Data - Microsoft Research (2013) 2017-12-30 we strive to develop a series of **new latent semantic models with a deep structure that project queries and documents into a common low-dimensional space** where the relevance of a document given a query is readily computed as the distance between them. The proposed deep structured semantic models are discriminatively trained by maximizing the conditional likelihood of the clicked documents given a query using the clickthrough data. To make our models applicable to large-scale Web search applications, we also use a technique called word hashing 2017-12-04 2017-12-04T09:05:41Z How is information coded in neural activity? (Quora) Topic Modeling with Scikit Learn – Aneesha Bakharia – Medium 2017-12-05T09:54:22Z 2017-12-05 2017-12-30 Combining word and entity embeddings for entity linking (ESWC 2017) The general approach for the entity linking task is to generate, for a given mention, a set of candidate entities from the base and, in a second step, determine which is the best one. This paper proposes a novel method for the second step which is based on the **joint learning of embeddings for the words in the text and the entities in the knowledge base**. 2017-12-30T01:14:53Z 2017-12-17T11:35:02Z 2017-12-17 “Had bitcoin been mined by doing something useful, then there would be a correspondence between useful work and the number of bitcoins you get … That creates a mental anchor point in people’s mind for how much a bitcoin should cost.” The Hard Math Behind Bitcoin's Global Warming Problem | WIRED Who pursues their goals with monomaniacal focus, oblivious to the possibility of negative consequences? 2017-12-19 2017-12-19T13:49:44Z Silicon Valley Is Turning Into Its Own Worst Fear > No one really knows how the most advanced algorithms do what they do. That could be a problem. The Dark Secret at the Heart of AI - MIT Technology Review 2017-12-31 2017-12-31T10:51:53Z How to Protect Yourself After the Next Big Corporate Hack | WIRED 2017-12-17T12:04:53Z 2017-12-17 2017-12-18 2017-12-18T00:50:52Z PAUL KRUGMAN: Bitcoin is a more obvious bubble than housing was 2017-12-12 2017-12-12T17:54:57Z D’anciens cadres de Facebook expriment leur « culpabilité » d’avoir contribué à son succès Machine Learning for Systems and Systems for Machine Learning (NIPS 2017) 2017-12-12T10:57:13Z 2017-12-12 Representation learning (in "Deep Learning", Ian Goodfellow and Yoshua Bengio and Aaron Courville) 2017-12-16T14:31:43Z 2017-12-16 2017-12-07 2017-12-07T23:35:09Z liveBook - Deep Learning with Python 2017-12-17 Have I been pwned? Check if your email has been compromised in a data breach 2017-12-17T12:02:04Z All About Eve 2017-12-18T23:06:12Z 2017-12-18 film de Mankiewicz. Une actrice de théâtre célèbre mais vieillissante (Bette Davis), et une jeunette qui prend sa place l2-normalize the dense vectors. machine learning - Text categorization: combining different kind of features - Data Science Stack Exchange 2017-12-06T16:51:37Z 2017-12-06 Edward Snowden’s New App Uses Your Smartphone to Physically Guard Your Laptop 2017-12-23T11:38:42Z 2017-12-23 2017-12-07 2017-12-07T00:38:02Z EDF : c’est pas bientôt fini le nucléaire ? 2017-12-16T17:56:28Z 2017-12-16 Why are palaeontologists suing Trump? | Elsa Panciroli | Science | The Guardian 2017-12-07 Quand la Terre était une boule de neige | CNRS Le journal 2017-12-07T22:49:44Z Burnistoun S1E1 - Voice Recognition Elevator - ELEVEN! - YouTube 2017-12-18T13:50:24Z 2017-12-18 Finding similar documents with Word2Vec and WMD (Word Mover’s Distance) 2017-12-23 2017-12-23T14:12:41Z gensim/WMD_tutorial.ipynb Exploring the Entire Tree of Life 2017-12-23 2017-12-23T11:09:13Z Lifemap LDA2vec: Word Embeddings in Topic Models – Towards Data Science 2017-12-11T13:46:53Z 2017-12-11 > While complex symbolic datasets often exhibit a latent hierarchical structure, state-of-the-art methods typically learn embeddings in Euclidean vector spaces, which do not account for this property. For this purpose, we introduce a new approach for learning hierarchical representations of symbolic data by embedding them into hyperbolic space [1705.08039] Poincaré Embeddings for Learning Hierarchical Representations 2017-12-16T14:41:31Z 2017-05-26T17:40:55Z 1705.08039 Douwe Kiela Maximilian Nickel Poincaré Embeddings for Learning Hierarchical Representations 2017-12-16 2017-05-22T23:14:36Z Maximilian Nickel Representation learning has become an invaluable approach for learning from symbolic data such as text and graphs. However, while complex symbolic datasets often exhibit a latent hierarchical structure, state-of-the-art methods typically learn embeddings in Euclidean vector spaces, which do not account for this property. For this purpose, we introduce a new approach for learning hierarchical representations of symbolic data by embedding them into hyperbolic space -- or more precisely into an n-dimensional Poincar\'e ball. Due to the underlying hyperbolic geometry, this allows us to learn parsimonious representations of symbolic data by simultaneously capturing hierarchy and similarity. We introduce an efficient algorithm to learn the embeddings based on Riemannian optimization and show experimentally that Poincar\'e embeddings outperform Euclidean embeddings significantly on data with latent hierarchies, both in terms of representation capacity and in terms of generalization ability. Xin-Yu Dai 2017-12-03 2015-06-28T16:17:40Z Latent Dirichlet Allocation (LDA) mining thematic structure of documents plays an important role in nature language processing and machine learning areas. However, the probability distribution from LDA only describes the statistical relationship of occurrences in the corpus and usually in practice, probability is not the best choice for feature representations. Recently, embedding methods have been proposed to represent words and documents by learning essential concepts and representations, such as Word2Vec and Doc2Vec. The embedded representations have shown more effectiveness than LDA-style representations in many tasks. In this paper, we propose the Topic2Vec approach which can learn topic representations in the same semantic vector space with words, as an alternative to probability. The experimental results show that Topic2Vec achieves interesting and meaningful results. Li-Qiang Niu [1506.08422] Topic2Vec: Learning Distributed Representations of Topics Topic2Vec: Learning Distributed Representations of Topics 2017-12-03T17:36:27Z Li-Qiang Niu Topic2Vec aims at learning topic representations along with word representations. Considering the simplicity and efficient solution, we just follow the optimization scheme that used in Word2Vec 1506.08422 2015-06-28T16:17:40Z Deep Learning (Ian Goodfellow and Yoshua Bengio and Aaron Courville) 2017-12-16 2017-12-16T14:25:02Z Advances in Pre-Training Distributed Word Representations Piotr Bojanowski Many Natural Language Processing applications nowadays rely on pre-trained word representations estimated from large text corpora such as news collections, Wikipedia and Web Crawl. In this paper, we show how to train high-quality word vector representations by using a combination of known tricks that are however rarely used together. The main result of our work is the new set of publicly available pre-trained models that outperform the current state of the art by a large margin on a number of tasks. Edouard Grave Armand Joulin 2017-12-29T20:52:48Z 2017-12-26T21:00:04Z Christian Puhrsch Tomas Mikolov > we show how to train high-quality word vector representations by using a combination of known tricks that are however rarely used together. The main result of our work is the new set of publicly available pre-trained models that outperform the current state of the art by a large margin on a number of tasks [1712.09405] Advances in Pre-Training Distributed Word Representations Tomas Mikolov 2017-12-29 1712.09405 2017-12-26T21:00:04Z 2017-12-12T11:00:44Z 2017-12-12 Deep Learning: Practice and Trends (NIPS 2017 Tutorial, parts I & II) - YouTube