Several multi-target regression methods were devel-oped in the last years aiming at improving predictive performanceby exploring inter-target correlation within the problem. However, none of these methods outperforms the others for all problems. This motivates the development of automatic approachesto recommend the most suitable multi-target regression method. In this paper, we propose a meta-learning system to recommend the best predictive method for a given multi-target regression problem. We performed experiments with a meta-dataset generated by a total of 648 synthetic datasets. These datasets were created to explore distinct inter-targets characteristics toward recommending the most promising method. In experiments, we evaluated four different algorithms with different biases as meta-learners. Our meta-dataset is composed of 58 meta-features, based on: statistical information, correlation characteristics, linear landmarking, from the distribution and smoothness of the data, and has four different meta-labels. Results showed that induced meta-models were able to recommend the best methodfor different base level datasets with a balanced accuracy superior to 70% using a Random Forest meta-model, which statistically outperformed the meta-learning baselines.

Contextual multi-armed bandit algorithms are widely used in sequential decision tasks such as news article recommendation systems, web page ad placement algorithms, and mobile health. Most of the existing algorithms have regret proportional to a polynomial function of the context dimension, $d$. In many applications however, it is often the case that contexts are high-dimensional with only a sparse subset of size $s_0 (\ll d)$ being correlated with the reward. We propose a novel algorithm, namely the Doubly-Robust Lasso Bandit algorithm, which exploits the sparse structure as in Lasso, while blending the doubly-robust technique used in missing data literature. The high-probability upper bound of the regret incurred by the proposed algorithm does not depend on the number of arms, has better dependency on $s_0$ than previous works, and scales with $\mathrm{log}(d)$ instead of a polynomial function of $d$. The proposed algorithm shows good performance when contexts of different arms are correlated and requires less tuning parameters than existing methods.

Numerous engineering problems of interest to the industry are often characterized by expensive black-box objective experiments or computer simulations. Obtaining insight into the problem or performing subsequent optimizations requires hundreds of thousands of evaluations of the objective function which is most often a practically unachievable task. Gaussian Process (GP) surrogate modeling replaces the expensive function with a cheap-to-evaluate data-driven probabilistic model. While the GP does not assume a functional form of the problem, it is defined by a set of parameters, called hyperparameters. The hyperparameters define the characteristics of the objective function, such as smoothness, magnitude, periodicity, etc. Accurately estimating these hyperparameters is a key ingredient in developing a reliable and generalizable surrogate model. Markov chain Monte Carlo (MCMC) is a ubiquitously used Bayesian method to estimate these hyperparameters. At the GE Global Research Center, a customized industry-strength Bayesian hybrid modeling framework utilizing the GP, called GEBHM, has been employed and validated over many years. GEBHM is very effective on problems of small and medium size, typically less than 1000 training points. However, the GP does not scale well in time with a growing dataset and problem dimensionality which can be a major impediment in such problems. In this work, we extend and implement in GEBHM an Adaptive Sequential Monte Carlo (ASMC) methodology for training the GP enabling the modeling of large-scale industry problems. This implementation saves computational time (especially for large-scale problems) while not sacrificing predictability over the current MCMC implementation. We demonstrate the effectiveness and accuracy of GEBHM with ASMC on four mathematical problems and on two challenging industry applications of varying complexity.

Optimization algorithms with momentum, e.g., Nesterov Accelerated Gradient and ADAM, have been widely used for building deep learning models because of their faster convergence rates compared to stochastic gradient descent (SGD). Momentum is a method that helps accelerate SGD in the relevant directions in variable updating, which can minify the oscillations of variables update route. Optimization algorithms with momentum usually allocate a fixed hyperparameter (e.g., \beta_1) as the weight of the momentum term. However, using a fixed weight is not applicable to some situations, and such a hyper-parameter can be extremely hard to tune in applications. In this paper, we will introduce a new optimization algorithm, namely DEAM (Discriminative wEight on Accumulated Momentum). Instead of assigning the momentum term with a fixed weight, DEAM proposes to compute the momentum weight in the learning process automatically. DEAM also involves a "backtrack" term, which can help accelerate the algorithm convergence by restricting redundant updates. Extensive experiments have been done on several real-world datasets. The experimental results demonstrate that DEAM can achieve a faster convergence rate than the existing optimization algorithms in training both the classic machine learning models and the recent deep learning models.

Domain generalization (DG) aims to incorporate knowledge from multiple source domains into a single model that could generalize well on unseen target domains. This problem is ubiquitous in practice since the distributions of the target data may rarely be identical to those of the source data. In this paper, we propose Multidomain Discriminant Analysis (MDA) to address DG of classification tasks in general situations. MDA learns a domain-invariant feature transformation that aims to achieve appealing properties, including a minimal divergence among domains within each class, a maximal separability among classes, and overall maximal compactness of all classes. Furthermore, we provide the bounds on excess risk and generalization error by learning theory analysis. Comprehensive experiments on synthetic and real benchmark datasets demonstrate the effectiveness of MDA.

The vision of automated driving is to increase both road safety and efficiency, while offering passengers a convenient travel experience. This requires that autonomous systems correctly estimate the current traffic scene and its likely evolution. In highway scenarios early recognition of cutin maneuvers is essential for risk-aware maneuver planning. In this paper, a statistical approach is proposed, which advantageously utilizes a set of prototypical lane change trajectories to realize both early maneuver detection and uncertainty-aware trajectory prediction for traffic participants. Generation of prototype trajectories from real traffic data is accomplished by Agglomerative Hierarchical Clustering. During clustering, the alignment of the cluster prototypes to each other is optimized and the cohesion of the resulting prototype is limited when two clusters merge. In the prediction stage, the similarity of observed vehicle motion and typical lane change patterns in the data base is evaluated to construct a set of significant features for maneuver classification via Boosted Decision T rees. The future trajectory is predicted combining typical lane change realizations in a mixture model. B-splines based trajectory adaptations guarantee continuity during transition from actually observed to predicted vehicle states. Quantitative evaluation results demonstrate the proposed concept's improved performance for both maneuver and trajectory prediction compared to a previously implemented reference approach. The development of automated driving functions is a central activity of industry and science.

On the path to establishing a global cybersecurity framework where each enterprise shares information about malicious behavior, an important question arises. How can a machine learning representation characterizing a cyber attack on one network be used to detect similar attacks on other enterprise networks if each networks has wildly different distributions of benign and malicious traffic? We address this issue by comparing the results of naively transferring a model across network domains and using CORrelation ALignment, to our novel adversarial Siamese neural network. Our proposed model learns attack representations that are more invariant to each network's particularities via an adversarial approach. It uses a simple ranking loss that prioritizes the labeling of the most egregious malicious events correctly over average accuracy. This is appropriate for driving an alert triage workflow wherein an analyst only has time to inspect the top few events ranked highest by the model. In terms of accuracy, the other approaches fail completely to detect any malicious events when models were trained on one dataset are evaluated on another for the first 100 events. While, the method presented here retrieves sizable proportions of malicious events, at the expense of some training instabilities due in adversarial modeling. We evaluate these approaches using 2 publicly available networking datasets, and suggest areas for future research.

In spite of the rapidly increasing number of applications of machine learning in various domains, a principled and systematic approach to the incorporation of domain knowledge in the engineering process is still lacking and ad hoc solutions that are difficult to validate are still the norm in practice, which is of growing concern not only in mission-critical applications. In this note, we introduce Probabilistic Approximate Logic (PALO) as a logic based on the notion of mean approximate probability to overcome conceptual and computational difficulties inherent to strictly probabilistic logics. The logic is approximate in several dimensions. Logical independence assumptions are used to obtain approximate probabilities, but by averaging over many instances of formulas a useful estimate of mean probability with known confidence can usually be obtained. To enable efficient computational inference, the logic has a continuous semantics that reflects only a subset of the structural properties of classical logic, but this imprecision can be partly compensated by richer theories obtained by classical inference or other means. Computational inference, which refers to the construction of models and validation of logical properties, is based on Stochastic Gradient Descent (SGD) and Markov Chain Monte Carlo (MCMC) techniques and hence another dimension where approximations are involved. We also present the Logical Imagination Engine (LIME), a prototypical implementation of PALO based on TensorFlow. Albeit not limited to the biological domain, we illustrate its operation in a quite substantial bioinformatics machine learning application concerned with network synthesis and analysis in a recent DARPA project.

Machine learning is an increasingly prevalent buzzword in the media. Its applications in science and the private sector are frequently discussed--but what about global development? Can it also help advance fields like health, agriculture, and financial inclusion? That's because it can help us uncover previously invisible patterns in data, to identify the most effective solutions and target them in the right way. Machine learning (ML) has been around for decades, but now is our chance to apply it to development challenges in new ways, for three reasons.

Minimizing a convex function of a measure with a sparsity-inducing penalty is a typical problem arising, e.g., in sparse spikes deconvolution or two-layer neural networks training. We show that this problem can be solved by discretizing the measure and running non-convex gradient descent on the positions and weights of the particles. For measures on a $d$-dimensional manifold and under some non-degeneracy assumptions, this leads to a global optimization algorithm with a complexity scaling as $\log(1/\epsilon)$ in the desired accuracy $\epsilon$, instead of $\epsilon^{-d}$ for convex methods. The key theoretical tools are a local convergence analysis in Wasserstein space and an analysis of a perturbed mirror descent in the space of measures. Our bounds involve quantities that are exponential in $d$ which is unavoidable under our assumptions.