The notions of concreteness and imageability, traditionally important in psycholinguistics, are gaining significance in semantic-oriented natural language processing tasks. In this paper we investigate the predictability of these two concepts via supervised learning, using word embeddings as explanatory variables. We perform predictions both within and across languages by exploiting collections of cross-lingual embeddings aligned to a single vector space. We show that the notions of concreteness and imageability are highly predictable both within and across languages, with a moderate loss of up to 20% in correlation when predicting across languages. We further show that the cross-lingual transfer via word embeddings is more efficient than the simple transfer via bilingual dictionaries.

Despite being very successful within the pattern recognition and machine learning community, graph-based methods are often unusable with many machine learning tools. This is because of the incompatibility of most of the mathematical operations in graph domain. Graph embedding has been proposed as a way to tackle these difficulties, which maps graphs to a vector space and makes the standard machine learning techniques applicable for them. However, it is well known that graph embedding techniques usually suffer from the loss of structural information. In this paper, given a graph, we consider its hierarchical structure for mapping it into a vector space. The hierarchical structure is constructed by topologically clustering the graph nodes, and considering each cluster as a node in the upper hierarchical level. Once this hierarchical structure of graph is constructed, we consider its various configurations of its parts, and use stochastic graphlet embedding (SGE) for mapping them into vector space. Broadly speaking, SGE produces a distribution of uniformly sampled low to high order graphlets as a way to embed graphs into the vector space. In what follows, the coarse-to-fine structure of a graph hierarchy and the statistics fetched through the distribution of low to high order stochastic graphlets complements each other and include important structural information with varied contexts. Altogether, these two techniques substantially cope with the usual information loss involved in graph embedding techniques, and it is not a surprise that we obtain more robust vector space embedding of graphs. This fact has been corroborated through a detailed experimental evaluation on various benchmark graph datasets, where we outperform the state-of-the-art methods.

Neural networks use their hidden layers to transform input data into linearly separable data clusters, with a linear or a perceptron type output layer making the final projection on the line perpendicular to the discriminating hyperplane. For complex data with multimodal distributions this transformation is difficult to learn. Projection on $k\geq 2$ line segments is the simplest extension of linear separability, defining much easier goal for the learning process. Simple problems are 2-separable, but problems with inherent complex logic may be solved in a simple way by $k$-separable projections. The difficulty of learning non-linear data distributions is shifted to separation of line intervals, simplifying the transformation of data by hidden network layers. For classification of difficult Boolean problems, such as the parity problem, linear projection combined with \ksep is sufficient and provides a powerful new target for learning. More complex targets may also be defined, changing the goal of learning from linear discrimination to creation of data distributions that can easily be handled by specialized models selected to analyze output distributions. This approach can replace many layers of transformation required by deep learning models.

The mean-median-mode trio involves the three most frequently used measures of central tendency of a dataset. They are taught within the very first classes of any course on basic Statistics. However, they do not share the same degree of importance: the sample mean (or average) is normally well understood and employed in everyday situations, the sample median is also useful and easy to visualize, but the mode, usually defined as the value of the dataset having the highest frequency of appearance, looks like a more bizarre measure of location. This uneven treatment was already noted by Dalenius (1965), but it keeps being present as of today, to some extent. Indeed, when the dataset consists of realizations from a continuous random variable then all the observed values are different with probability one and, therefore, the mode does not even make much sense.

Bayesian optimization is an approach to optimizing objective functions that take a long time (minutes or hours) to evaluate. It is best-suited for optimization over continuous domains of less than 20 dimensions, and tolerates stochastic noise in function evaluations. It builds a surrogate for the objective and quantifies the uncertainty in that surrogate using a Bayesian machine learning technique, Gaussian process regression, and then uses an acquisition function defined from this surrogate to decide where to sample. In this tutorial, we describe how Bayesian optimization works, including Gaussian process regression and three common acquisition functions: expected improvement, entropy search, and knowledge gradient. We then discuss more advanced techniques, including running multiple function evaluations in parallel, multi-fidelity and multi-information source optimization, expensive-to-evaluate constraints, random environmental conditions, multi-task Bayesian optimization, and the inclusion of derivative information. We conclude with a discussion of Bayesian optimization software and future research directions in the field. Within our tutorial material we provide a generalization of expected improvement to noisy evaluations, beyond the noise-free setting where it is more commonly applied. This generalization is justified by a formal decision-theoretic argument, standing in contrast to previous ad hoc modifications.

Model interpretability is an increasingly important component of practical machine learning. Some of the most common forms of interpretability systems are example-based, local, and global explanations. One of the main challenges in interpretability is designing explanation systems that can capture aspects of each of these explanation types, in order to develop a more thorough understanding of the model. We address this challenge in a novel model called SLIM that uses local linear modeling techniques along with a dual interpretation of random forests (both as a supervised neighborhood approach and as a feature selection method). SLIM has two fundamental advantages over existing interpretability systems. First, while it is effective as a black-box explanation system, SLIM itself is a highly accurate predictive model that provides faithful self explanations, and thus sidesteps the typical accuracy-interpretability trade-off. Second, SLIM provides both example- based and local explanations and can detect global patterns, which allows it to diagnose limitations in its local explanations.

The key issue of few-shot learning is learning to generalize. In this paper, we propose a large margin principle to improve the generalization capacity of metric based methods for few-shot learning. To realize it, we develop a unified framework to learn a more discriminative metric space by augmenting the softmax classification loss function with a large margin distance loss function for training. Extensive experiments on two state-of-the-art few-shot learning models, graph neural networks and prototypical networks, show that our method can improve the performance of existing models substantially with very little computational overhead, demonstrating the effectiveness of the large margin principle and the potential of our method.

We propose a method to infer domain-specific models such as classifiers for unseen domains, from which no data are given in the training phase, without domain semantic descriptors. When training and test distributions are different, standard supervised learning methods perform poorly. Zero-shot domain adaptation attempts to alleviate this problem by inferring models that generalize well to unseen domains by using training data in multiple source domains. Existing methods use observed semantic descriptors characterizing domains such as time information to infer the domain-specific models for the unseen domains. However, it cannot always be assumed that such metadata can be used in real-world applications. The proposed method can infer appropriate domain-specific models without any semantic descriptors by introducing the concept of latent domain vectors, which are latent representations for the domains and are used for inferring the models. The latent domain vector for the unseen domain is inferred from the set of the feature vectors in the corresponding domain, which is given in the testing phase. The domain-specific models consist of two components: the first is for extracting a representation of a feature vector to be predicted, and the second is for inferring model parameters given the latent domain vector. The posterior distributions of the latent domain vectors and the domain-specific models are parametrized by neural networks, and are optimized by maximizing the variational lower bound using stochastic gradient descent. The effectiveness of the proposed method was demonstrated through experiments using one regression and two classification tasks.

Lecture 3 continues our discussion of linear classifiers. We introduce the idea of a loss function to quantify our unhappiness with a model's predictions, and discuss two commonly used loss functions for image classification: the multiclass SVM loss and the multinomial logistic regression loss. We introduce the idea of regularization as a mechanism to fight overfitting, with weight decay as a concrete example. We introduce the idea of optimization and the stochastic gradient descent algorithm. We also briefly discuss the use of feature representations in computer vision.

This is part one of a three-part tutorial series in which you will use R to perform a variety of analytic tasks on a case study of musical lyrics by the legendary artist, Prince. Musical lyrics may represent an artist's perspective, but popular songs reveal what society wants to hear. Lyric analysis is no easy task. Because it is often structured so differently than prose, it requires caution with assumptions and a uniquely discriminant choice of analytic techniques. Musical lyrics permeate our lives and influence our thoughts with subtle ubiquity. The concept of Predictive Lyrics is beginning to buzz and is more prevalent as a subject of research papers and graduate theses. This case study will just touch on a few pieces of this emerging subject. To celebrate the inspiring and diverse body of work left behind by Prince, you will explore the sometimes obvious, but often hidden, messages in his lyrics. However, you don't have to like Prince's music to appreciate the influence he had on the development of many genres globally. Rolling Stone magazine listed Prince as the 18th best songwriter of all time, just behind the likes of Bob Dylan, John Lennon, Paul Simon, Joni Mitchell and Stevie Wonder. Lyric analysis is slowly finding its way into data science communities as the possibility of predicting "Hit Songs" approaches reality. Prince was a man bursting with music - a wildly prolific songwriter, a virtuoso on guitars, keyboards and drums and a master architect of funk, rock, R&B and pop, even as his music defied genres. In this tutorial, Part One of the series, you'll utilize text mining techniques on a set of lyrics using the tidy text framework.