Complex networks are often categorized according to the underlying phenomena that they represent such as molecular interactions, re-tweets, and brain activity. In this work, we investigate the problem of predicting the category (domain) of arbitrary networks. This includes complex networks from different domains as well as synthetically generated graphs from five different network models. A classification accuracy of $96.6\%$ is achieved using a random forest classifier with both real and synthetic networks. This work makes two important findings. First, our results indicate that complex networks from various domains have distinct structural properties that allow us to predict with high accuracy the category of a new previously unseen network. Second, synthetic graphs are trivial to classify as the classification model can predict with near-certainty the network model used to generate it. Overall, the results demonstrate that networks drawn from different domains (and network models) are trivial to distinguish using only a handful of simple structural properties.

In just a few years, deep learning-powered AI and other forms of machine learning have exploded from niche tools into the underlying basis of nearly every major modern online platform. Yet, even as these algorithms increasingly wield near-absolute power over what we see and say online, we have precious little visibility into how they function. If AI tools can robustly prevent hate speech and terrorist recruiting, flag false news and delete financial scams, they will be a welcome addition to our online lives. On the other hand, without any visibility into how often they get things wrong, we have little reason to trust their successes. AI and machine learning have become ubiquitous on the modern web, powering everything from security scanning to content moderation.

We consider a Bayesian framework for estimating a high-dimensional sparse precision matrix, in which adaptive shrinkage and sparsity are induced by a mixture of Laplace priors. Besides discussing our formulation from the Bayesian standpoint, we investigate the MAP (maximum a posteriori) estimator from a penalized likelihood perspective that gives rise to a new non-convex penalty approximating the $\ell_0$ penalty. Optimal error rates for estimation consistency in terms of various matrix norms along with selection consistency for sparse structure recovery are shown for the unique MAP estimator under mild conditions. For fast and efficient computation, an EM algorithm is proposed to compute the MAP estimator of the precision matrix and (approximate) posterior probabilities on the edges of the underlying sparse structure. Through extensive simulation studies and a real application to a call center data, we have demonstrated the fine performance of our method compared with existing alternatives.

At the end of each summer for the last 14 years, the small Welsh town of Porthcawl has been invaded. Every year its 16,000 population is swamped by up to 35,000 Elvis fans. Many people attending the yearly festival look the same: they slick back their hair, throw on oversized sunglasses and don white flares. At 2017's Elvis festival, impersonators were faced with something different. Police were trialling automated facial recognition technology to track down criminals.

Learning tasks on source code (i.e., formal languages) have been considered recently, but most work has tried to transfer natural language methods and does not capitalize on the unique opportunities offered by code's known syntax. For example, long-range dependencies induced by using the same variable or function in distant locations are often not considered. We propose to use graphs to represent both the syntactic and semantic structure of code and use graph-based deep learning methods to learn to reason over program structures. In this work, we present how to construct graphs from source code and how to scale Gated Graph Neural Networks training to such large graphs. We evaluate our method on two tasks: VarNaming, in which a network attempts to predict the name of a variable given its usage, and VarMisuse, in which the network learns to reason about selecting the correct variable that should be used at a given program location. Our comparison to methods that use less structured program representations shows the advantages of modeling known structure, and suggests that our models learn to infer meaningful names and to solve the VarMisuse task in many cases. Additionally, our testing showed that VarMisuse identifies a number of bugs in mature open-source projects.

When he testified before Congress last month, Facebook CEO Mark Zuckerberg discussed the problem of using artificial intelligence to identify online hate speech. He said he was optimistic that in five to 10 years, "We will have AI tools that can get into some of the linguistic nuances of different types of content to be more accurate in flagging content for our systems, but today we're not just there on that." Brittan Heller (@brittanheller) is director of the Anti-Defamation League's Center for Technology and Society and works with social media companies to reduce cyberhate and online harassment. As an expert on hate speech who recently developed an AI-based system to study online hate, I can confidently say that Zuckerberg is both right and wrong. He is right that AI is not a panacea, since hate speech relies on nuances that algorithms cannot fully detect. At the same time, just because AI does not solve the problem entirely doesn't mean it's useless.

One thing I find weird is that we have lots of discussion of deep learning in complex detection and recognition tasks, but very few people talk about how useful deep learning can be for simple but time consuming image data processing tasks, particularly in medical research. In this post I spend a bit of time cleaning up the CXR14 dataset, and in 4 hours find 430 images with various problems that shouldn't be in the dataset (a csv identifying these images is included in the post). While the prevalence of these problems is super low ( 50/100,000), since the visual challenge is very easy the models can achieve absurdly low false positive rates. I even get an AUROC of 1.0 in a 2000 image validation set on one task:) In doing so, cleaning this dataset to remove 3 different problems didn't take me weeks to pore through each image, but under a day. Certainly nothing in the post is technically groundbreaking, but it is hopefully a prompt to consider deep learning when you are doing time consuming processing.

Semi-supervised bootstrapping techniques for relationship extraction from text iteratively expand a set of initial seed instances. Due to the lack of labeled data, a key challenge in bootstrapping is semantic drift: if a false positive instance is added during an iteration, then all following iterations are contaminated. We introduce BREX, a new bootstrapping method that protects against such contamination by highly effective confidence assessment. This is achieved by using entity and template seeds jointly (as opposed to just one as in previous work), by expanding entities and templates in parallel and in a mutually constraining fashion in each iteration and by introducing higherquality similarity measures for templates. Experimental results show that BREX achieves an F1 that is 0.13 (0.87 vs. 0.74) better than the state of the art for four relationships.

Machine learning (ML) is all about getting machines to learn but how do we know how well they are doing? Suppose we have existing data about people who buy books from our website – age, amount spent, preferred author and so on. These columns of data are known, in ML terms, as the "Predictors". For some of the customers we also know their gender, for others we don't, but we want to know it for all customers. We start with the customers where we do know the gender and we feed both the Predictor and Response data into a machine learning algorithm and get it to build a model that can (hopefully) forecast the Response from the Predictors.

We develop, discuss, and compare several inference techniques to constrain theory parameters in collider experiments. By harnessing the latent-space structure of particle physics processes, we extract extra information from the simulator. This augmented data can be used to train neural networks that precisely estimate the likelihood ratio. The new methods scale well to many observables and high-dimensional parameter spaces, do not require any approximations of the parton shower and detector response, and can be evaluated in microseconds. Using weak-boson-fusion Higgs production as an example process, we compare the performance of several techniques. The best results are found for likelihood ratio estimators trained with extra information about the score, the gradient of the log likelihood function with respect to the theory parameters. The score also provides sufficient statistics that contain all the information needed for inference in the neighborhood of the Standard Model. These methods enable us to put significantly stronger bounds on effective dimension-six operators than the traditional approach based on histograms. They also outperform generic machine learning methods that do not make use of the particle physics structure, demonstrating their potential to substantially improve the new physics reach of the LHC legacy results.