Generative models that learn to associate variations in the output along isolated attributes with coordinate directions in latent space are said to possess disentangled representations. This has been a recent topic of great interest, but remains poorly understood. We show that even for GANs that do not possess disentangled representations, one can find paths in latent space over which local disentanglement occurs. These paths are determined by the leading right-singular vectors of the Jacobian of the generator with respect to its input. We describe an efficient regularizer that aligns these vectors with the coordinate axes, and show that it can be used to induce high-quality, disentangled representations in GANs, in a completely unsupervised manner.

In complex transfer learning scenarios new tasks might not be tightly linked to previous tasks. Approaches that transfer information contained only in the final parameters of a source model will therefore struggle. Instead, transfer learning at a higher level of abstraction is needed. We propose Leap, a framework that achieves this by transferring knowledge across learning processes. We associate each task with a manifold on which the training process travels from initialization to final parameters and construct a meta learning objective that minimizes the expected length of this path. Our framework leverages only information obtained during training and can be computed on the fly at negligible cost. We demonstrate that our framework outperforms competing methods, both in meta learning and transfer learning, on a set of computer vision tasks. Finally, we demonstrate that Leap can transfer knowledge across learning processes in demanding Reinforcement learning environments (Atari) that involve millions of gradient steps.

In this paper, we compare predictive models for students' final performance in a blended course using a set of generic features collected from the first six weeks of class. These features were extracted from students' online homework submission logs as well as other online actions. We compare the effectiveness of 5 different ML algorithms (SVMs, Support Vector Regression, Decision Tree, Naive Bayes and K-Nearest Neighbor). We found that SVMs outperform other models and improve when compared to the baseline. This study demonstrates feasible implementations for predictive models that rely on common data from blended courses that can be used to monitor students' progress and to tailor instruction.

Lets start with this simple number series question. Find the next few numbers in this series 2, 5, 8, 11, 14... From the numbers it is trivial to guess that each number is 3 more than the previous number. So the numbers would be 17, 20, 23. This would be the end of this post if there were no machine learning. Can we predict the next number using some machine learning techniques.

In "Why Use K-Means for Time Series Data? (Part One)," I give an overview of how to use different statistical functions and K-Means Clustering for anomaly detection for time series data. I recommend checking that out if you're unfamiliar with either. I am borrowing the code and dataset for this portion from Amid Fish's tutorial. Please take a look at it, it's pretty awesome. In this example, I will show you how you can detect anomalies in EKG data via contextual anomaly detection with K-Means Clustering.

The softmax function has been primarily employed in reinforcement learning (RL) to improve exploration and provide a differentiable approximation to the max function, as also observed in the mellowmax paper by Asadi and Littman. This paper instead focuses on using the softmax function in the Bellman updates, independent of the exploration strategy. Our main theory provides a performance bound for the softmax Bellman operator, and shows it converges to the standard Bellman operator exponentially fast in the inverse temperature parameter. We also prove that under certain conditions, the softmax operator can reduce the overestimation error and the gradient noise. A detailed comparison among different Bellman operators is then presented to show the trade-off when selecting them. We apply the softmax operator to deep RL by combining it with the deep Q-network (DQN) and double DQN algorithms in an off-policy fashion, and demonstrate that these variants can often achieve better performance in several Atari games, and compare favorably to their mellowmax counterpart.

Feature selection, which searches for the most representative features in observed data, is critical for health data analysis. Unlike feature extraction, such as PCA and autoencoder based methods, feature selection preserves interpretability, meaning that the selected features provide direct information about certain health conditions (i.e., the label). Thus, feature selection allows domain experts, such as clinicians, to understand the predictions made by machine learning based systems, as well as improve their own diagnostic skills. Mutual information is often used as a basis for feature selection since it measures dependencies between features and labels. In this paper, we introduce a novel mutual information based feature selection (MIBFS) method called SURI, which boosts features with high unique relevant information. We compare SURI to existing MIBFS methods using 3 different classifiers on 6 publicly available healthcare data sets. The results indicate that, in addition to preserving interpretability, SURI selects more relevant feature subsets which lead to higher classification performance. More importantly, we explore the dynamics of mutual information on a public low-dimensional health data set via exhaustive search. The results suggest the important role of unique relevant information in feature selection and verify the principles behind SURI.

Medical imaging machine learning algorithms are usually evaluated on a single dataset. Although training and testing are performed on different subsets of the dataset, models built on one study show limited capability to generalize to other studies. While database bias has been recognized as a serious problem in the computer vision community, it has remained largely unnoticed in medical imaging research. Transfer learning thus remains confined to the re-use of feature representations requiring re-training on the new dataset. As a result, machine learning models do not generalize even when trained on imaging datasets that were captured to study the same variable of interest. The ability to transfer knowledge gleaned from one study to another, without the need for re-training, if possible, would provide reassurance that the models are learning knowledge fundamental to the problem under study instead of latching onto the idiosyncracies of a dataset. In this paper, we situate the problem of dataset bias in the context of medical imaging studies. We show empirical evidence that such a problem exists in medical datasets. We then present a framework to unlearn study membership as a means to handle the problem of database bias. Our main idea is to take the data from the original feature space to an intermediate space where the data points are indistinguishable in terms of which study they come from, while maintaining the recognition capability with respect to the variable of interest. This will promote models which learn the more general properties of the etiology under study instead of aligning to dataset-specific peculiarities. Essentially, our proposed model learns to unlearn the dataset bias.

Missing data is a common problem in real-world settings and particularly relevant in healthcare applications where researchers use Electronic Health Records (EHR) and results of observational studies to apply analytics methods. This issue becomes even more prominent for longitudinal data sets, where multiple instances of the same individual correspond to different observations in time. Standard imputation methods do not take into account patient specific information incorporated in multivariate panel data. We introduce the novel imputation algorithm MedImpute that addresses this problem, extending the flexible framework of OptImpute suggested by Bertsimas et al. (2018). Our algorithm provides imputations for data sets with missing continuous and categorical features, and we present the formulation and implement scalable first-order methods for a $K$-NN model. We test the performance of our algorithm on longitudinal data from the Framingham Heart Study when data are missing completely at random (MCAR). We demonstrate that MedImpute leads to significant improvements in both imputation accuracy and downstream model AUC compared to state-of-the-art methods.

Labeling training data is one of the most costly bottlenecks in developing or modifying machine learning-based applications. We survey how resources from across an organization can be used as weak supervision sources for three classification tasks at Google, in order to bring development time and cost down by an order of magnitude. We build on the Snorkel framework, extending it as a new system, Snorkel DryBell, which integrates with Google's distributed production systems and enables engineers to develop and execute weak supervision strategies over millions of examples in less than thirty minutes. We find that Snorkel DryBell creates classifiers of comparable quality to ones trained using up to tens of thousands of hand-labeled examples, in part by leveraging organizational resources not servable in production which contribute an average 52% performance improvement to the weakly supervised classifiers.