I come from the Cambridge machine learning group. More than once I heard people referring to us as "the most Bayesian machine learning group in the world". I mean, we do work with probabilistic models and uncertainty on a daily basis. Maybe that's why it felt so weird playing with those deep learning models (I know, joining the party very late). You see, I spent the last several years working mostly with Gaussian processes, modelling probability distributions over functions. I'm used to uncertainty bounds for decision making, in a similar way many biologists rely on model uncertainty to analyse their data. Working with point estimates alone felt weird to me. I couldn't tell whether the new model I was playing with was making sensible predictions or just guessing at random. I'm certain you've come across this problem yourself, either analysing data or solving some tasks, where you wished you could tell whether your model is certain about its output, asking yourself "maybe I need to use more diverse data? or perhaps change the model?". Most deep learning tools operate in a very different setting to the probabilistic models which possess this invaluable uncertainty information, as one would believe. I recently spent some time trying to understand why these deep learning models work so well – trying to relate them to new research from the last couple of years. I was quite surprised to see how close these were to my beloved Gaussian processes. I was even more surprised to see that we can get uncertainty information from these deep learning models for free – without changing a thing. Update (29/09/2015): I spotted a typo in the calculation of \tau; this has been fixed below.

In this work, we particularly focus on the complex relationship between land-use and transport offering an innovative approach to the problem by using land-use features at two differing levels of granularity (the more general land-use sector types and the more granular amenity structures) to evaluate their impact on public transit ridership in both time and space. To quantify the interdependencies, we explored three machine learning models and demonstrate that the decision tree model performs best in terms of overall performance--good predictive accuracy, generality, computational efficiency, and "interpretability". We then demonstrate how the developed framework can be applied to urban planning for transit-oriented development by exploring practicable scenarios based on Singapore's urban plan toward 2030, which includes the development of "regional centers" (RCs) across the city-state. This trend, on the other hand, eventually reverses (particularly during peak hours) with continued strategic increase in amenities; a tipping point at 55% increase is identified where ridership begins to decline and at 110%, the predicted ridership begins to fall below current levels.

I was initially writing on a different topic (related to analytics). I had almost finished writing it. I had put in about 2 hours and written an average article. If I had made it live, it would have done OK! But something in me stopped me from making it live.

If TensorFlow is your primary framework, and you are looking for a simple & high-level model definition interface to make your life easier, this tutorial is for you. Keras layers and models are fully compatible with pure-TensorFlow tensors, and as a result, Keras makes a great model definition add-on for TensorFlow, and can even be used alongside other TensorFlow libraries. Note that this tutorial assumes that you have configured Keras to use the TensorFlow backend (instead of Theano). Here are instructions on how to do this. Let's start with a simple example: MNIST digits classification.

Various land-use configurations are known to have wide-ranging effects on the dynamics of and within other city components including the transportation system. In this work, we particularly focus on the complex relationship between land-use and transport offering an innovative approach to the problem by using land-use features at two differing levels of granularity (the more general land-use sector types and the more granular amenity structures) to evaluate their impact on public transit ridership in both time and space. To quantify the interdependencies, we explored three machine learning models and demonstrate that the decision tree model performs best in terms of overall performance--good predictive accuracy, generality, computational efficiency, and "interpretability". Results also reveal that amenity-related features are better predictors than the more general ones, which suggests that high-resolution geo-information can provide more insights into the dependence of transit ridership on land-use. We then demonstrate how the developed framework can be applied to urban planning for transit-oriented development by exploring practicable scenarios based on Singapore's urban plan toward 2030, which includes the development of "regional centers" (RCs) across the city-state.

Respondents were asked what they saw as the biggest area of opportunity for Big Data in comparison to traditional systems, with 62% agreeing that they consider real time analysis as the biggest area of opportunity today. "It's not long ago we were visiting enterprises and having to explain why they should look at big data. Today in 2016, Big Data Analytics is already considered a necessity to remain competitive by 63% of organizations," explains Serge Haziyev, VP Technology Services, SoftServe. "It's very encouraging that machine learning has featured so prominently in this survey. I find that businesses that take the plunge and implement machine learning techniques realize the benefits early on – it's a big step forward because it delivers prescriptive insights enabling businesses to not only understand what customers are doing, but why."

The new team has a distinct goal: to invent the future of Search, a voice-activated, human-like entity that can answer any query intelligently. "We are building the ultimate assistant. In two years, you can expect Google to become a personal life assistant across multiple surfaces, including your phone, Google Home, even cars," Mogenet said. Some of Google's best-known products are already shaped by machine learning, the ability of computers to spot patterns in large datasets and learn by example. For instance, Google Photos uses it to understand the content of an image.

We have been hearing predictions for decades of a takeover of the world by artificial intelligence. In 1957, Herbert A. Simon predicted that within 10 years a digital computer would be the world's chess champion. That didn't happen until 1996. And despite Marvin Minsky's 1970 prediction that "in from three to eight years we will have a machine with the general intelligence of an average human being," we still consider that a feat of science fiction. The pioneers of artificial intelligence were surely off on the timing, but they weren't wrong; AI is coming.

In his quest to transform the way the Pentagon wages war, Defense Secretary Ashton B. Carter has turned to Silicon Valley, hoping its experimental culture, innovation and sense of urgency would rub off on the rigid bureaucracy he runs. Carter has made several trips to the Valley and appointed Eric Schmidt, the chairman of Google's parent company to an advisory board. And recently he sat down at the Pentagon with Elon Musk to see what suggestions the billionaire founder of Tesla and SpaceX might have to make the nation's military more efficient and daring. "Having an incentive structure that rewards innovation is extremely important," he said in an interview after the meeting. Whatever you reward will happen." The Pentagon finds itself in a new arms race, struggling to keep pace with forms of combat that are fought with bytes as well as bullets. The technological advancements disrupting established business sectors are now shaking up the world of war -- where robots, swarming drones ...

When most people think of artificial intelligence (AI) they think of HAL 9000 from "2001: A Space Odyssey," Data from "Star Trek," or more recently, the android Ava from "Ex Machina." But to a computer scientist that isn't what AI necessarily is, and the question "what is AI?" can be a complicated one. One of the standard textbooks in the field, by University of California computer scientists Stuart Russell and Google's director of research, Peter Norvig, puts artificial intelligence in to four broad categories: The differences between them can be subtle, notes Ernest Davis, a professor of computer science at New York University. AlphaGo, the computer program that beat a world champion at Go, acts rationally when it plays the game (it plays to win). But it doesn't necessarily think the way a human being does, though it engages in some of the same pattern-recognition tasks.