Continuing from where we left, we delve deeper into how to develop machine learning (ML) algorithms using TensorFlow from a scikit-learn developer's perspective. If you'd like to know the reasons to move to TensorFlow, motivations, do read my earlier post for Reasons to move to TensorFlow and a simple classification program that highlights similarities of developing for scikit-learn and TensorFlow. In the earlier post, we compared the fit and predict paradigm similarities in scikit-learn and TensorFlow. In this post, I want to show we can develop a TensorFlow classification framework with Scikit-learn's data processing and reporting tools. This will give a good method to interweave both the frameworks to come up with a neat and concise framework.

In this article we propose a novel ranking algorithm, referred to as HierLPR, for the multi-label classification problem when the candidate labels follow a known hierarchical structure. HierLPR is motivated by a new metric called eAUC that we design to assess the ranking of classification decisions. This metric, associated with the hit curve and local precision rate, emphasizes the accuracy of the first calls. We show that HierLPR optimizes eAUC under the tree constraint and some light assumptions on the dependency between the nodes in the hierarchy. We also provide a strategy to make calls for each node based on the ordering produced by HierLPR, with the intent of controlling FDR or maximizing F-score. The performance of our proposed methods is demonstrated on synthetic datasets as well as a real example of disease diagnosis using NCBI GEO datasets. In these cases, HierLPR shows a favorable result over competing methods in the early part of the precision-recall curve.

Diabetic eye disease is one of the fastest growing causes of preventable blindness. With the advent of anti-VEGF (vascular endothelial growth factor) therapies, it has become increasingly important to detect center-involved diabetic macular edema. However, center-involved diabetic macular edema is diagnosed using optical coherence tomography (OCT), which is not generally available at screening sites because of cost and workflow constraints. Instead, screening programs rely on the detection of hard exudates as a proxy for DME on color fundus photographs, often resulting in high false positive or false negative calls. To improve the accuracy of DME screening, we trained a deep learning model to use color fundus photographs to predict DME grades derived from OCT exams. Our "OCT-DME" model had an AUC of 0.89 (95% CI: 0.87-0.91), which corresponds to a sensitivity of 85% at a specificity of 80%. In comparison, three retinal specialists had similar sensitivities (82-85%), but only half the specificity (45-50%, p<0.001 for each comparison with model). The positive predictive value (PPV) of the OCT-DME model was 61% (95% CI: 56-66%), approximately double the 36-38% by the retina specialists. In addition, we used saliency and other techniques to examine how the model is making its prediction. The ability of deep learning algorithms to make clinically relevant predictions that generally require sophisticated 3D-imaging equipment from simple 2D images has broad relevance to many other applications in medical imaging.

In this paper, we present a new explainability formalism to make clear the impact of each variable on the predictions given by black-box decision rules. Our method consists in evaluating the decision rules on test samples generated in such a way that each variable is stressed incrementally while preserving the original distribution of the machine learning problem. We then propose a new computation-ally efficient algorithm to stress the variables, which only reweights the reference observations and predictions. This makes our methodology scalable to large datasets. Results obtained on standard machine learning datasets are presented and discussed.

We propose robust sparse reduced rank regression and robust sparse principal component analysis for analyzing large and complex high-dimensional data with heavy-tailed random noise. The proposed methods are based on convex relaxations of rank-and sparsity-constrained non-convex optimization problems, which are solved using the alternating direction method of multipliers (ADMM) algorithm. For robust sparse reduced rank regression, we establish non-asymptotic estimation error bounds under both Frobenius and nuclear norms, while existing results focus mostly on rank-selection and prediction consistency. Our theoretical results quantify the tradeoff between heavy-tailedness of the random noise and statistical bias. For random noise with bounded $(1+\delta)$th moment with $\delta \in (0,1)$, the rate of convergence is a function of $\delta$, and is slower than the sub-Gaussian-type deviation bounds; for random noise with bounded second moment, we recover the results obtained under sub-Gaussian noise. Furthermore, the transition between the two regimes is smooth. For robust sparse principal component analysis, we propose to truncate the observed data, and show that this truncation will lead to consistent estimation of the eigenvectors. We then establish theoretical results similar to those of robust sparse reduced rank regression. We illustrate the performance of these methods via extensive numerical studies and two real data applications.

Addressing fairness in machine learning models has recently attracted a lot of attention, as it will ensure continued confidence of the general public in the deployment of machine learning systems. Here, we focus on mitigating harm of a biased system that offers much better quality outputs for certain groups than for others. We show that bias in the output can naturally be handled in Gaussian process classification (GPC) models by introducing a latent target output that will modulate the likelihood function. This simple formulation has several advantages: first, it is a unified framework for several notions of fairness (demographic parity, equalized odds, and equal opportunity); second, it allows encoding our knowledge of what the bias in outputs should be; and third, it can be solved by using off-the-shelf GPC packages.

Automatic understanding of domain specific texts in order to extract useful relationships for later use is a non-trivial task. One such relationship would be between railroad accidents' causes and their correspondent descriptions in reports. From 2001 to 2016 rail accidents in the U.S. cost more than $4.6B. Railroads involved in accidents are required to submit an accident report to the Federal Railroad Administration (FRA). These reports contain a variety of fixed field entries including primary cause of the accidents (a coded variable with 389 values) as well as a narrative field which is a short text description of the accident. Although these narratives provide more information than a fixed field entry, the terminologies used in these reports are not easy to understand by a non-expert reader. Therefore, providing an assisting method to fill in the primary cause from such domain specific texts(narratives) would help to label the accidents with more accuracy. Another important question for transportation safety is whether the reported accident cause is consistent with narrative description. To address these questions, we applied deep learning methods together with powerful word embeddings such as Word2Vec and GloVe to classify accident cause values for the primary cause field using the text in the narratives. The results show that such approaches can both accurately classify accident causes based on report narratives and find important inconsistencies in accident reporting.

The UCR Time Series Archive - introduced in 2002, has become an important resource in the time series data mining community, with at least one thousand published papers making use of at least one dataset from the archive. The original incarnation of the archive had sixteen datasets but since that time, it has gone through periodic expansions. The last expansion took place in the summer of 2015 when the archive grew from 45 datasets to 85 datasets. This paper introduces and will focus on the new data expansion from 85 to 128 datasets. Beyond expanding this valuable resource, this paper offers pragmatic advice to anyone who may wish to evaluate a new algorithm on the archive. Finally, this paper makes a novel and yet actionable claim: of the hundreds of papers that show an improvement over the standard baseline (1-Nearest Neighbor classification), a large fraction may be misattributing the reasons for their improvement. Moreover, they may have been able to achieve the same improvement with a much simpler modification, requiring just a single line of code.

This research tested the following well known strategies to deal with binary imbalanced data on 82 different real life data sets (sampled to imbalance rates of 5%, 3%, 1%, and 0.1%): class weight, SMOTE, Underbagging, and a baseline (just the base classifier). As base classifiers we used SVM with RBF kernel, random forests, and gradient boosting machines and we measured the quality of the resulting classifier using 6 different metrics (Area under the curve, Accuracy, F-measure, G-mean, Matthew's correlation coefficient and Balanced accuracy). The best strategy strongly depends on the metric used to measure the quality of the classifier. For AUC and accuracy class weight and the baseline perform better; for F-measure and MCC, SMOTE performs better; and for G-mean and balanced accuracy, underbagging.

Advances in supervised learning have enabled accurate prediction in biological systems governed by complex interactions among biomolecules. However, state-of-the-art predictive algorithms are typically black-boxes, learning statistical interactions that are difficult to translate into testable hypotheses. The iterative Random Forest algorithm took a step towards bridging this gap by providing a computationally tractable procedure to identify the stable, high-order feature interactions that drive the predictive accuracy of Random Forests (RF). Here we refine the interactions identified by iRF to explicitly map responses as a function of interacting features. Our method, signed iRF, describes subsets of rules that frequently occur on RF decision paths. We refer to these rule subsets as signed interactions. Signed interactions share not only the same set of interacting features but also exhibit similar thresholding behavior, and thus describe a consistent functional relationship between interacting features and responses. We describe stable and predictive importance metrics to rank signed interactions. For each SPIM, we define null importance metrics that characterize its expected behavior under known structure. We evaluate our proposed approach in biologically inspired simulations and two case studies: predicting enhancer activity and spatial gene expression patterns. In the case of enhancer activity, s-iRF recovers one of the few experimentally validated high-order interactions and suggests novel enhancer elements where this interaction may be active. In the case of spatial gene expression patterns, s-iRF recovers all 11 reported links in the gap gene network. By refining the process of interaction recovery, our approach has the potential to guide mechanistic inquiry into systems whose scale and complexity is beyond human comprehension.