Hierarchical Classification (HC) is a supervised learning problem where unlabeled instances are classified into a taxonomy of classes. Several methods that utilize the hierarchical structure have been developed to improve the HC performance. However, in most cases apriori defined hierarchical structure by domain experts is inconsistent; as a consequence performance improvement is not noticeable in comparison to flat classification methods. We propose a scalable data-driven filter based rewiring approach to modify an expert-defined hierarchy. Experimental comparisons of top-down HC with our modified hierarchy, on a wide range of datasets shows classification performance improvement over the baseline hierarchy (i:e:, defined by expert), clustered hierarchy and flattening based hierarchy modification approaches. In comparison to existing rewiring approaches, our developed method (rewHier) is computationally efficient, enabling it to scale to datasets with large numbers of classes, instances and features. We also show that our modified hierarchy leads to improved classification performance for classes with few training samples in comparison to flat and state-of-the-art HC approaches.

High dimensionality is a frequently confronted problem in many applications of machine learning. It increases time and space requirements for processing the data. Moreover, many machine learning methods tend to over-fit and become less interpretable in the presence of many irrelevant or redundant features. A common way to address this problem is the dimensionality reduction. Principal Component Analysis (PCA) (Jolliffe, 2002) is probably the most widely used dimensionality reduction method. However, it is an unsupervised dimensionality reduction method and does not consider the labels of the data. In order to take the label information into account, supervised dimensionality reduction methods are favored. Linear Discriminant Analysis (LDA) (Anderson, 1968), which is initially proposed as a classification method, is an important supervised dimensionality reduction method.

A new computationally efficient dependence measure, and an adaptive statistical test of independence, are proposed. The dependence measure is the difference between analytic embeddings of the joint distribution and the product of the marginals, evaluated at a finite set of locations (features). These features are chosen so as to maximize a lower bound on the test power, resulting in a test that is data-efficient, and that runs in linear time (with respect to the sample size n). The optimized features can be interpreted as evidence to reject the null hypothesis, indicating regions in the joint domain where the joint distribution and the product of the marginals differ most. Consistency of the independence test is established, for an appropriate choice of features. In real-world benchmarks, independence tests using the optimized features perform comparably to the state-of-the-art quadratic-time HSIC test, and outperform competing O(n) and O(n log n) tests.

We propose a unified and systematic framework for performing online nonnegative matrix factorization in the presence of outliers. Our framework is particularly suited to large-scale data. We propose two solvers based on projected gradient descent and the alternating direction method of multipliers. We prove that the sequence of objective values converges almost surely by appealing to the quasi-martingale convergence theorem. We also show the sequence of learned dictionaries converges to the set of stationary points of the expected loss function almost surely. In addition, we extend our basic problem formulation to various settings with different constraints and regularizers. We also adapt the solvers and analyses to each setting. We perform extensive experiments on both synthetic and real datasets. These experiments demonstrate the computational efficiency and efficacy of our algorithms on tasks such as (parts-based) basis learning, image denoising, shadow removal and foreground-background separation.

In this paper, we consider clustering data that is assumed to come from one of finitely many pointed convex polyhedral cones. This model is referred to as the Union of Polyhedral Cones (UOPC) model. Similar to the Union of Subspaces (UOS) model where each data from each subspace is generated from a (unknown) basis, in the UOPC model each data from each cone is assumed to be generated from a finite number of (unknown) \emph{extreme rays}.To cluster data under this model, we consider several algorithms - (a) Sparse Subspace Clustering by Non-negative constraints Lasso (NCL), (b) Least squares approximation (LSA), and (c) K-nearest neighbor (KNN) algorithm to arrive at affinity between data points. Spectral Clustering (SC) is then applied on the resulting affinity matrix to cluster data into different polyhedral cones. We show that on an average KNN outperforms both NCL and LSA and for this algorithm we provide the deterministic conditions for correct clustering. For an affinity measure between the cones it is shown that as long as the cones are not very coherent and as long as the density of data within each cone exceeds a threshold, KNN leads to accurate clustering. Finally, simulation results on real datasets (MNIST and YaleFace datasets) depict that the proposed algorithm works well on real data indicating the utility of the UOPC model and the proposed algorithm.

Years of evolution got something right. Scientists at the University of Tokyo found that, in trying to keep their humanoid robot cool, the most effective way to avoid overheating was to make it sweat. Kengoro is made up of 108 motors, with a frame laser sintered from aluminum, and researchers found that making the robot sweat was an effective cooling system for a bot filled to the brim with bolts. Even with the space-saving technique, the bot weighs 123 pounds and stands 5 feet 7 inches tall. "Usually the frame of a robot is only used to support forces," lead author Toyotaka Kozuki told IEEE Spectrum in an interview published Friday.

The DeepMind artificial intelligence (AI) being developed by Google's parent company, Alphabet, can now intelligently build on what's already inside its memory, the system's programmers have announced. Their new hybrid system – called a Differential Neural Computer (DNC) – pairs a neural network with the vast data storage of conventional computers, and the AI is smart enough to navigate and learn from this external data bank. What the DNC is doing is effectively combining external memory (like the external hard drive where all your photos get stored) with the neural network approach of AI, where a massive number of interconnected nodes work dynamically to simulate a brain. "These models... can learn from examples like neural networks, but they can also store complex data like computers," write DeepMind researchers Alexander Graves and Greg Wayne in a blog post. At the heart of the DNC is a controller that constantly optimises its responses, comparing its results with the desired and correct ones.

In the literature there are multiple machine learning techniques that have been used successfully in clinical data analysis. However, there is little information about the parameter configurations, the required data transformations to prepare the data used to train and evaluate the models and the impact of these decisions in the accuracy of the predictive model. This research tackles these issues, using the clinical data of MIMICII to build features from physiological measure patterns to predict the decease of patients inside the hospital in the next 24 hours, building predictive models based on Logistic Regression, Neural Networks, Decision Trees and Nearest Neighbors. In particular, we use data associated to physiological measures of 3220 patients, where 2385 left the hospital alive and 835 passed in the hospital. The results show that the chosen strategy for building features from physiological data gives good results with Neural Networks and Logistic Regression with radial kernel models and the parameter configuration plays a fundamental role in the models performance.

We are entering an extremely critical time in history where society will change dramatically – how we work, live and play. Science fiction is morphing into reality. Flying cars exist, cars that drive themselves are on the road, and artificial intelligence that automates our lives is here. And you? what do you think about?

The currently ongoing Seizure Prediction competition--hosted by Melbourne University AES, MathWorks, and NIH--invites Kagglers to accurately forecast the occurrence of seizures using intracranial EEG recordings. This competition uniquely focuses on seizure prediction using long-term electrical brain activity from human patients obtained from the world first clinical trial of the implantable NeuroVista Seizure Advisory Sytem. In this blog post, you'll learn about the contest's potential to positively impact the lives of those who suffer from epilepsy, outcomes of previous seizure prediction contests on Kaggle, as well as resources which will help you get started in the competition including a free temporary MATLAB license and starter code. This competition is sponsored by MathWorks, the National Institutes of Health (NINDS), the American Epilepsy Society and the University of Melbourne, and organised in partnership with the Alliance for Epilepsy Research, the University of Pennsylvania and the Mayo Clinic. For many people with epilepsy, seizures reoccur at random times and greatly disrupt their cognitive and emotional state, their ability to work and drive, and their social and economic situation.