Local representation vs distributed representation l Local representation each concept is represented by one symbol e.g. Giraff 1, Panda 2, Lion 3, Tiger 4 no interfere, noise immunity, precise l Distributed representation each concept is represented by a set of symbol, and each symbol participates in representing many concepts Generalizable less accurate interfere Giraff Pand Lion Tiger Long neck four legs body hair paw pad 61. High dimensional vector vs low dimensional data l High dimensional vector Random two vectors are always almost orthogonal many concepts can be stored within one vector u w x y z, Same characteristics as local representation l Low dimensional vector Interfere each other Cannot keep precise memory Beneficial for generalization l Interference and generalization are strongly related 62. Two layer feedforward network memory augmented network [Vaswani 17] l Memory augmented network a V Softmax(Kq) K is a key matrix (i-th row corresponds to a key for i-th memory) V is a value matix.

Instead of worrying so much about robots taking away jobs, maybe we should worry more about wages being too low for robots to even get a chance. Seasonal labor for harvesting agricultural products, particularly fruits and vegetables, is dependent on human labor from a diminishing universe of willing workers. Robots that can supplement or replace human workers in the harvesting process are being developed and tested in startups and academia, but almost all are not yet ready for prime time. In a NY Times article written by Neil Irwin entitled Rethinking Low Productivity, productivity growth has been on a downward path since the Financial Crisis. Irwin, who writes about economic trends, asks whether the downward trend is the cause of low growth, or the result, a troubling question in the dynamics of the agriculture industry.

Top Machine Learning & Data Mining Books - in this post, we have scraped various signals (e.g. A highly rated book on Amazon written by a well-known author Christopher M. Bishop who is a distinguished Scientist at Microsoft Research in Cambridge where he leads the Machine Learning and Perception group. The "Machine Learning" is a well-know book in the field of Machine Learning written by Tom Mitchell - an American computer scientist professor from the Carnegie Mellon University. This foundational text is a comprehensive introduction to statistical natural language processing (NLP).

Two Texas men who were killed earlier this year may have been lured to their deaths by an online dating app used by the pair's killers, investigators said Monday. Harris County sheriff's investigators said the app, which was not named, might have been used to entice Glenser Soliman, 44, and An Vinh Nguyen, 26, The Houston Chronicle reported. Soliman, a nurse at St. Luke's Medical Center, was found dead a few miles from his residence on Feb. 25 after being declared missing on Feb. 16. Nguyen, a student studying hotel and restaurant management at the University of Houston, has not been seen since March 31. Deputies believe the student is dead, but his body has not been found.

Churn prediction is big business. It minimizes customer defection by predicting which customers are likely to cancel a subscription to a service. Though originally used within the telecommunications industry, it has become common practice across banks, ISPs, insurance firms, and other verticals. The prediction process is heavily data-driven and often utilizes advanced machine learning techniques. In this post, we'll take a look at what types of customer data are typically used, do some preliminary analysis of the data, and generate churn prediction models -- all with Spark and its machine learning frameworks.

Say Machine Learning to someone, and if they recognize the term, they'll probably think, "tech company." But while the origin stories of transformative technologies like machine learning, deep learning, and artificial intelligence often seem to take root in Silicon Valley, the truth is these are industry-agnostic innovations. Their impact is being felt across countless fields you might never have thought of as being ripe for technological advancement. Think about it like this: If you were a farmer, and someone came to you and said, there's a technology out there that can accurately predict your crop yields, would you be interested? Well, this is exactly what Descartes Labs does.

Have you ever felt guilty about leaving your dog home alone, wishing there was a way to calm it? Amazon's Audible may have the answer - a good book. Dog behavior expert Cesar Millan has teamed up with the firm to create audiobooks for dogs, which use human voices to tell stories to dogs when nobody is at home, keeping then company. Dog behavior expert Cesar Millan (pictured) teamed up with audible to conduct their own study to see what impact audiobooks have on dogs when their owners are away. By using a speaker along with one of the audiobooks, users can play one of a range of audible's audiobooks, with recommended titles including: According to a demonstration video by Millan, having to be away from our dogs can be a stressful situation, joined by guilt and shame for having to be away from our pets for long periods of time, for example when going to work.

Online review communities are dynamic as users join and leave, adopt new vocabulary, and adapt to evolving trends. Recent work has shown that recommender systems benefit from explicit consideration of user experience. However, prior work assumes a fixed number of discrete experience levels, whereas in reality users gain experience and mature continuously over time. This paper presents a new model that captures the continuous evolution of user experience, and the resulting language model in reviews and other posts. Our model is unsupervised and combines principles of Geometric Brownian Motion, Brownian Motion, and Latent Dirichlet Allocation to trace a smooth temporal progression of user experience and language model respectively. We develop practical algorithms for estimating the model parameters from data and for inference with our model (e.g., to recommend items). Extensive experiments with five real-world datasets show that our model not only fits data better than discrete-model baselines, but also outperforms state-of-the-art methods for predicting item ratings.

Embarrassingly (communication-free) parallel Markov chain Monte Carlo (MCMC) methods are commonly used in learning graphical models. However, MCMC cannot be directly applied in learning topic models because of the quasi-ergodicity problem caused by multimodal distribution of topics. In this paper, we develop an embarrassingly parallel MCMC algorithm for sLDA. Our algorithm works by switching the order of sampled topics combination and labeling variable prediction in sLDA, in which it overcomes the quasi-ergodicity problem because high-dimension topics that follow a multimodal distribution are projected into one-dimension document labels that follow a unimodal distribution. Our empirical experiments confirm that the out-of-sample prediction performance using our embarrassingly parallel algorithm is comparable to non-parallel sLDA while the computation time is significantly reduced.

Applied statisticians use sequential regression procedures to produce a ranking of explanatory variables and, in settings of low correlations between variables and strong true effect sizes, expect that variables at the very top of this ranking are true. In a regime of certain sparsity levels, however, three examples of sequential procedures---forward stepwise, the lasso, and least angle regression---are shown to include the first spurious variable unexpectedly early. We derive a rigorous, sharp prediction of the rank of the first spurious variable for the three procedures, demonstrating that the first spurious variable occurs earlier and earlier as the regression coefficients get denser. This counterintuitive phenomenon persists for independent Gaussian random designs and an arbitrarily large magnitude of the true effects. We further gain a better understanding of the phenomenon by identifying the underlying cause and then leverage the insights to introduce a simple visualization tool termed the "double-ranking diagram" to improve on sequential methods. As a byproduct of these findings, we obtain the first provable result certifying the exact equivalence between the lasso and least angle regression in the early stages of solution paths beyond orthogonal designs. This equivalence can seamlessly carry over many important model selection results concerning the lasso to least angle regression.