The fear of the Earth being struck by an enormous asteroid has been the source for many science fiction films, but the threat of mankind being wiped out by an impact may be lower than we think. Until now, scientists used the strength of rocks found on Earth to model the damage these giant space boulders would cause. But researchers have now sacrificed a rare sample of meteorites to test the physical properties of asteroids and found they crumbled more easily than expected. It is an idea popular in science fiction, but the threat of mankind being wiped out by asteroids (artist's impression pictured) might be a lot lower than you think. Even though space rocks fall to Earth as meteorites all the time, very few are recovered.

Most Machine Learning courses are given from the perspective of a researcher/academic and focus on the theory and mathematics of the machine learning models. This workshop takes the perspective of learning by working on real machine learning problems using open source tools and platforms. We'll go all the way from data preparation to the integration of predictive models in applications and their deployment in production. "Just like development where you don't need to know a thing about computability or big-O notation to write code and ship useful and reliable software, you can work machine learning problems end-to-end without a background in statistics, probability and linear algebra." The workshop is agnostic and features the best open source Python libraries (Pandas, scikit-learn, SKLL), APIs and ML-as-a-Service platforms (Microsoft Azure ML & Cortana Intelligence Suite, Amazon ML, BigML) for developers getting started in Machine Learning.

Worldwide spending on cybersecurity likely topped 75 billion last year, researchers at Gartner estimated, with companies more wary than ever of the risks posed by data breaches and other digital attacks. And along with rising costs, the sheer volume of digital security data has also increased dramatically: IBM estimated in a recent study that the average organization sees more than 200,000 pieces of security event data per day and that more than 10,000 security-related research papers are published every year. "Security researchers are getting hit with a firehose," says Caleb Barlow, vice president of IBM Security. "Once they get done with today, they've got another deluge of data coming tomorrow." To help companies handle that flood of data, IBM says it's training its Watson artificial intelligence platform--previously known for using its natural language processing power to beat humans on Jeopardy--to parse cybersecurity information, from automated network-level threat reports to blog posts from security professionals. According to Barlow, the company hopes to train the system to detect and understand threats to computer systems and to answer questions from human security professionals about incidents they detect on their networks.

Second-order optimization methods such as natural gradient descent have the potential to speed up training of neural networks by correcting for the curvature of the loss function. Unfortunately, the exact natural gradient is impractical to compute for large models, and most approximations either require an expensive iterative procedure or make crude approximations to the curvature. We present Kronecker Factors for Convolution (KFC), a tractable approximation to the Fisher matrix for convolutional networks based on a structured probabilistic model for the distribution over backpropagated derivatives. Similarly to the recently proposed Kronecker-Factored Approximate Curvature (K-FAC), each block of the approximate Fisher matrix decomposes as the Kronecker product of small matrices, allowing for efficient inversion. KFC captures important curvature information while still yielding comparably efficient updates to stochastic gradient descent (SGD). We show that the updates are invariant to commonly used reparameterizations, such as centering of the activations. In our experiments, approximate natural gradient descent with KFC was able to train convolutional networks several times faster than carefully tuned SGD. Furthermore, it was able to train the networks in 10-20 times fewer iterations than SGD, suggesting its potential applicability in a distributed setting.

Each human genome is a 3 billion base pair set of encoding instructions. Decoding the genome using deep learning fundamentally differs from most tasks, as we do not know the full structure of the data and therefore cannot design architectures to suit it. As such, architectures that fit the structure of genomics should be learned not prescribed. Here, we develop a novel search algorithm, applicable across domains, that discovers an optimal architecture which simultaneously learns general genomic patterns and identifies the most important sequence motifs in predicting functional genomic outcomes. The architectures we find using this algorithm succeed at using only RNA expression data to predict gene regulatory structure, learn human-interpretable visualizations of key sequence motifs, and surpass state-of-the-art results on benchmark genomics challenges.

Abstract--A central problem in analyzing networks is partitioning them into modules or communities. One of the best tools for this is the stochastic block model, which clusters vertices into blocks with statistically homogeneous pattern of links. Despite its flexibility and popularity, there has been a lack of principled statistical model selection criteria for the stochastic block model. Here we propose a Bayesian framework for choosing the number of blocks as well as comparing it to the more elaborate degree-corrected block models, ultimately leading to a universal model selection framework capable of comparing multiple modeling combinations. We will also investigate its connection to the minimum description length principle. I NTRODUCTION An important task in network analysis is community detection, or finding groups of similar vertices which can then be analyzed separately [1]. Community structures offer clues to the processes which generated the graph, on scales ranging from face-to-face social interaction [2] through social-media communications [3] to the organization of food webs [4]. However, previous work often defines a "community" as a group of vertices with high density of connections within the group and a low density of connections to the rest of the network. While this type of assortative community structure is generally the case in social networks, we are interested in a more general definition of functional community--a group of vertices that connect to the rest of the network in similar ways. A set of similar predators form a functional group in a food web, not because they eat each other, but because they feed on similar prey.

We study an online learning framework introduced by Mannor and Shamir (2011) in which the feedback is specified by a graph, in a setting where the graph may vary from round to round and is \emph{never fully revealed} to the learner. We show a large gap between the adversarial and the stochastic cases. In the adversarial case, we prove that even for dense feedback graphs, the learner cannot improve upon a trivial regret bound obtained by ignoring any additional feedback besides her own loss. In contrast, in the stochastic case we give an algorithm that achieves $\widetilde \Theta(\sqrt{\alpha T})$ regret over $T$ rounds, provided that the independence numbers of the hidden feedback graphs are at most $\alpha$. We also extend our results to a more general feedback model, in which the learner does not necessarily observe her own loss, and show that, even in simple cases, concealing the feedback graphs might render a learnable problem unlearnable.

We analyze stochastic algorithms for optimizing nonconvex, nonsmooth finite-sum problems, where the nonconvex part is smooth and the nonsmooth part is convex. Surprisingly, unlike the smooth case, our knowledge of this fundamental problem is very limited. For example, it is not known whether the proximal stochastic gradient method with constant minibatch converges to a stationary point. To tackle this issue, we develop fast stochastic algorithms that provably converge to a stationary point for constant minibatches. Furthermore, using a variant of these algorithms, we show provably faster convergence than batch proximal gradient descent. Finally, we prove global linear convergence rate for an interesting subclass of nonsmooth nonconvex functions, that subsumes several recent works. This paper builds upon our recent series of papers on fast stochastic methods for smooth nonconvex optimization [22, 23], with a novel analysis for nonconvex and nonsmooth functions.

There is resurging interest, in statistics and machine learning, in solvers for ordinary differential equations (ODEs) that return probability measures instead of point estimates. Recently, Conrad et al. introduced a sampling-based class of methods that are 'well-calibrated' in a specific sense. But the computational cost of these methods is significantly above that of classic methods. On the other hand, Schober et al. pointed out a precise connection between classic Runge-Kutta ODE solvers and Gaussian filters, which gives only a rough probabilistic calibration, but at negligible cost overhead. By formulating the solution of ODEs as approximate inference in linear Gaussian SDEs, we investigate a range of probabilistic ODE solvers, that bridge the trade-off between computational cost and probabilistic calibration, and identify the inaccurate gradient measurement as the crucial source of uncertainty. We propose the novel filtering-based method Bayesian Quadrature filtering (BQF) which uses Bayesian quadrature to actively learn the imprecision in the gradient measurement by collecting multiple gradient evaluations.

Supplying job schedulers with information on how long the jobs are expected to run enabled the development of the backfilling algorithms, which leverage this information to pack the jobs more efficiently and improve system utilization [1]. These algorithms, however, were designed for parallel systems, in which the jobs require many processors in order to execute, and processor fragmentation (idleness) is a big concern. In those environments the scheduler needs to know the actual runtimes of the jobs (use numeric predictions) to be able to optimize the schedule and improve performance [10]. Our work targets systems in which most jobs are serial, like server farms that are used for software testing. In those environments sophisticated scheduling algorithms are not required, and in order to improve performance it is enough to simply separate the short jobs from the long and assign them to different queues in the system [12]. This separation reduces the likelihood that short jobs will be delayed after long ones, improves the average turnaround times of the jobs and overall system throughput.