This is the second installment in a three-part review of 2016 in machine learning and deep learning. Part One, here, covered general trends. In Part Two, we review the year in open source machine learning and deep learning projects. Part Three will cover commercial machine learning and deep learning software and services. There are thousands of open source projects on the market today, and we cannot cover them all. We've selected the most relevant projects based on usage reported in surveys of data scientists, as well as development activity recorded in OpenHub. In this post, we limit the scope to projects with a non-profit governance structure, and those offered by commercial ventures that do not also provide licensed software. Part Three will include software vendors who offer open source "community" editions together with commercially licensed software.

This algorithm forms the basis for many modern day ML algorithms, most notably neural networks. In addition, we'll discuss the perceptron algorithm's cousin, logistic regression. And then we'll conclude with an introduction to SVMs, or support vector machines, which are perhaps one of the most flexible algorithms used today. In machine learning, there are two general classes of algorithms. You'll remember that in our last post we discussed regression and classification.

Autonomous vehicles are already making profound choices about whose lives matter, according to experts, so we might want to pay attention. "Every time the car makes a complex manoeuvre, it is implicitly making trade-off in terms of risks to different parties," Iyad Rahwan, an MIT cognitive scientist, wrote in an email. The most well-known issues in AV ethics are trolly problems -- moral questions dating back to the era of trollies that ask whose lives should be sacrificed in an unavoidable crash. For instance, if a person falls onto the road in front of a fast-moving AV, and the car can either swerve into a traffic barrier, potentially killing the passenger, or go straight, potentially killing the pedestrian, what should it do? Rahwan and colleagues have studied what humans consider the moral action in no-win scenarios (you can judge your own cases at their crowd-sourced project, Moral Machine).

Autoregression is a time series model that uses observations from previous time steps as input to a regression equation to predict the value at the next time step. It is a very simple idea that can result in accurate forecasts on a range of time series problems. In this tutorial, you will discover how to implement an autoregressive model for time series forecasting with Python. Autoregression Models for Time Series Forecasting With Python Photo by Umberto Salvagnin, some rights reserved. A regression model, such as linear regression, models an output value based on a linear combination of input values.

On the west coast of Australia, Amanda Hodgson is launching drones out towards the Indian Ocean so that they can photograph the water from above. The photos are a way of locating dugongs, or sea cows, in the bay near Perth--part of an effort to prevent the extinction of these endangered marine mammals. The trouble is that Hodgson and her team don't have the time needed to examine all those aerial photos. There are too many of them--about 45,000--and spotting the dugongs is far too difficult for the untrained eye. Deep learning is remaking Google, Facebook, Microsoft, and Amazon.

You're whizzing along a dark, outback highway when the symptoms start. Your mouth is so dry it's getting hard to swallow or talk, your grip on the steering wheel is weakening and that centre white line just turned into two. You pull in at a small town hospital where the tired junior doctor tells you it's probably a virus, and then hands your case over to night staff. She retakes your history wearing Google Glass, and the web-enabled headgear uses voice recognition to input your symptoms into a massive data base. It's botulism; very rare, often fatal, and you're just in time to get the antitoxin.

One crucial goal in kernel online learning is to bound the model size. Common approaches employ budget maintenance procedures to restrict the model sizes using removal, projection, or merging strategies. Although projection and merging, in the literature, are known to be the most effective strategies, they demand extensive computation whilst removal strategy fails to retain information of the removed vectors. An alternative way to address the model size problem is to apply random features to approximate the kernel function. This allows the model to be maintained directly in the random feature space, hence effectively resolve the curse of kernelization. However, this approach still suffers from a serious shortcoming as it needs to use a high dimensional random feature space to achieve a sufficiently accurate kernel approximation. Consequently, it leads to a significant increase in the computational cost. To address all of these aforementioned challenges, we present in this paper the Dual Space Gradient Descent (DualSGD), a novel framework that utilizes random features as an auxiliary space to maintain information from data points removed during budget maintenance. Consequently, our approach permits the budget to be maintained in a simple, direct and elegant way while simultaneously mitigating the impact of the dimensionality issue on learning performance. We further provide convergence analysis and extensively conduct experiments on five real-world datasets to demonstrate the predictive performance and scalability of our proposed method in comparison with the state-of-the-art baselines.

While neural machine translation (NMT) is making good progress in the past two years, tens of millions of bilingual sentence pairs are needed for its training. However, human labeling is very costly. To tackle this training data bottleneck, we develop a dual-learning mechanism, which can enable an NMT system to automatically learn from unlabeled data through a dual-learning game. This mechanism is inspired by the following observation: any machine translation task has a dual task, e.g., English-to-French translation (primal) versus French-to-English translation (dual); the primal and dual tasks can form a closed loop, and generate informative feedback signals to train the translation models, even if without the involvement of a human labeler. In the dual-learning mechanism, we use one agent to represent the model for the primal task and the other agent to represent the model for the dual task, then ask them to teach each other through a reinforcement learning process. Based on the feedback signals generated during this process (e.g., the language-model likelihood of the output of a model, and the reconstruction error of the original sentence after the primal and dual translations), we can iteratively update the two models until convergence (e.g., using the policy gradient methods). We call the corresponding approach to neural machine translation \emph{dual-NMT}. Experiments show that dual-NMT works very well on English$\leftrightarrow$French translation; especially, by learning from monolingual data (with 10\% bilingual data for warm start), it achieves a comparable accuracy to NMT trained from the full bilingual data for the French-to-English translation task.

Apple has stayed true to its promise and published its first academic paper on artificial intelligence. The world's most valuable company has traditionally kept its AI research private but earlier this month Ruslan Salakhutdinov, director of AI research at Apple, made a pledge to start being more open. The new Apple paper -- published December 22 and titled "Learning from simulated and unsupervised images through adversarial training" -- gives an insight into some of the techniques that Apple is using to develop AI. In the study, which was published through the Cornell University Library, Apple researchers explain a technique that can be used to improve how an algorithm learns to "see" what is in an image. The paper's six authors state that using synthetic images (such as those seen in a video game), as opposed to real-world images, can be more efficient when it comes to training AI models known as neural networks, which are designed to think in the same way as the human brain. Because synthetic image data is already labelled and annotated while real-world images aren't.

In 2016, the White House recognized the importance of AI at its Frontiers Conference. The concept of driverless cars became a reality, with Uber's self-driving fleet in Pittsburgh and Tesla's new models equipped with the hardware for full autonomy. Google's DeepMind platform, AlphaGo, beat the world champion of the game--10 years ahead of predictions. "Increasing use of machine learning and knowledge-based modeling methods" are major trends to watch in 2017, said Marie desJardins, associate dean and professor of computer science at the University of Maryland, Baltimore County. How will this play out?