Content creation has long been regarded as one of the most challenging obstacles to personalized learning. In recent years, however, online platforms have managed to mobilize both audiences and content creators in large numbers, creating new opportunities to revisit the pursuit of personalization at scale. We describe initial results from a real-world implementation of a system for algorithmically generating learning paths at Udemy.com, a two-sided online educational marketplace with over 150,000 courses and over 50 million users. Our initial investigations suggest the potential effectiveness of automated approaches for guiding self-directed learners toward courses that help them achieve their desired learning outcomes.

MOTIVATION: Advances in high-resolution microscopy have recently made possible the analysis of gene expression at the level of individual cells. The fixed lineage of cells in the adult worm Caenorhabditis elegans makes this organism an ideal model for studying complex biological processes like development and aging. However, annotating individual cells in images of adult C.elegans typically requires expertise and significant manual effort. Automation of this task is therefore critical to enabling high-resolution studies of a large number of genes. RESULTS: In this article, we describe an automated method for annotating a subset of 154 cells (including various muscle, intestinal and hypodermal cells) in high-resolution images of adult C.elegans. We formulate the task of labeling cells within an image as a combinatorial optimization problem, where the goal is to minimize a scoring function that compares cells in a test input image with cells from a training atlas of manually annotated worms according to various spatial and morphological characteristics. We propose an approach for solving this problem based on reduction to minimum-cost maximum-flow and apply a cross-entropy-based learning algorithm to tune the weights of our scoring function. We achieve 84% median accuracy across a set of 154 cell labels in this highly variable system. These results demonstrate the feasibility of the automatic annotation of microscopy-based images in adult C.elegans.

Many learning algorithms rely on the curvature (in particular, strong convexity) of regularized objective functions to provide good theoretical performance guarantees. In practice, the choice of regularization penalty that gives the best testing set performance may result in objective functions with little or even no curvature. In these cases, algorithms designed specifically for regularized objectives often either fail completely or require some modification that involves a substantial compromise in performance.

We present a general Bayesian framework for hyperparameter tuning in L2-regularized supervised learning models. Paradoxically, our algorithm works by first analytically integrating out the hyperparameters from the model. We find a local optimum of the resulting non-convex optimization problem efficiently using a majorization-minimization (MM) algorithm, in which the non-convex problem is reduced to a series of convex L2-regularized parameter estimation tasks. The principal appeal of our method is its simplicity: the updates for choosing the L2-regularized subproblems in each step are trivial to implement (or even perform by hand), and each subproblem can be efficiently solved by adapting existing solvers. Empirical results on a variety of supervised learning models show that our algorithm is competitive with both grid-search and gradient-based algorithms, but is more efficient and far easier to implement.

MOTIVATION: Genome-wide association studies are commonly used to identify possible associations between genetic variations and diseases. These studies mainly focus on identifying individual single nucleotide polymorphisms (SNPs) potentially linked with one disease of interest. In this work, we introduce a novel methodology that identifies similarities between diseases using information from a large number of SNPs. We separate the diseases for which we have individual genotype data into one reference disease and several query diseases. We train a classifier that distinguishes between individuals that have the reference disease and a set of control individuals. This classifier is then used to classify the individuals that have the query diseases. We can then rank query diseases according to the average classification of the individuals in each disease set, and identify which of the query diseases are more similar to the reference disease. We repeat these classification and comparison steps so that each disease is used once as reference disease. RESULTS: We apply this approach using a decision tree classifier to the genotype data of seven common diseases and two shared control sets provided by the Wellcome Trust Case Control Consortium. We show that this approach identifies the known genetic similarity between type 1 diabetes and rheumatoid arthritis, and identifies a new putative similarity between bipolar disease and hypertension.

MOTIVATION: The need for accurate and efficient tools for computational RNA structure analysis has become increasingly apparent over the last several years: RNA folding algorithms underlie numerous applications in bioinformatics, ranging from microarray probe selection to de novo non-coding RNA gene prediction. In this work, we present RAF (RNA Alignment and Folding), an efficient algorithm for simultaneous alignment and consensus folding of unaligned RNA sequences. Algorithmically, RAF exploits sparsity in the set of likely pairing and alignment candidates for each nucleotide (as identified by the CONTRAfold or CONTRAlign programs) to achieve an effectively quadratic running time for simultaneous pairwise alignment and folding. RAF's fast sparse dynamic programming, in turn, serves as the inference engine within a discriminative machine learning algorithm for parameter estimation. RESULTS: In cross-validated benchmark tests, RAF achieves accuracies equaling or surpassing the current best approaches for RNA multiple sequence secondary structure prediction. However, RAF requires nearly an order of magnitude less time than other simultaneous folding and alignment methods, thus making it especially appropriate for high-throughput studies. AVAILABILITY: Source code for RAF is available at:http://contra.stanford.edu/contrafold/.

Large-margin structured estimation methods work by minimizing a convex upper bound of loss functions. While they allow for efficient optimization algorithms, these convex formulations are not tight and sacrifice the ability to accurately model the true loss. We present tighter non-convex bounds based on generalizing the notion of a ramp loss from binary classification to structured estimation. We show that a small modification of existing optimization algorithms suffices to solve this modified problem. On structured prediction tasks such as protein sequence alignment and web page ranking, our algorithm leads to improved accuracy.

Using multiple regularization hyperparameters is an effective method for managing model complexity in problems where input features have varying amounts of noise. While algorithms for choosing multiple hyperparameters are often used in neural networks and support vector machines, they are not common in structured prediction tasks, such as sequence labeling or parsing. In this paper, we consider the problem of learning regularization hyperparameters for log-linear models, a class of probabilistic models for structured prediction tasks which includes conditional random fields (CRFs). Using an implicit differentiation trick, we derive an efficient gradient-based method for learning Gaussian regularization priors with multiple hyperparameters. In both simulations and the real-world task of computational RNA secondary structure prediction, we find that multiple hyperparameter learning provides a significant boost in accuracy compared to models learned using only a single regularization hyperparameter.

We consider the problem of training a conditional random field (CRF) to maximize per-label predictive accuracy on a training set, an approach motivated by the principle of empirical risk minimization. We give a gradient-based procedure for minimizing an arbitrarily accurate approximation of the empirical risk under a Hamming loss function. In experiments with both simulated and real data, our optimization procedure gives significantly better testing performance than several current approaches for CRF training, especially in situations of high label noise.

Linear text classification algorithms work by computing an inner prod- uct between a test document vector and a parameter vector. In many such algorithms, including naive Bayes and most TFIDF variants, the parame- ters are determined by some simple, closed-form, function of training set statistics; we call this mapping mapping from statistics to parameters, the parameter function. Much research in text classification over the last few decades has consisted of manual efforts to identify better parameter func- tions. In this paper, we propose an algorithm for automatically learning this function from related classification problems. The parameter func- tion found by our algorithm then defines a new learning algorithm for text classification, which we can apply to novel classification tasks. We find that our learned classifier outperforms existing methods on a variety of multiclass text classification tasks. arg maxk2f1;:::;KgPn