Production recommendation systems rely on embedding methods to represent various features. An impeding challenge in practice is that the large embedding matrix incurs substantial memory footprint in serving as the number of features grows over time. We propose a similarity-aware embedding matrix compression method called Saec to address this challenge. Saec clusters similar features within a field to reduce the embedding matrix size. Saec also adopts a fast clustering optimization based on feature frequency to drastically improve clustering time. We implement and evaluate Saec on Numerous, the production distributed machine learning system in Tencent, with 10-day worth of feature data from QQ mobile browser. Testbed experiments show that Saec reduces the number of embedding vectors by two orders of magnitude, compresses the embedding size by ~27x, and delivers the same AUC and log loss performance.

In many applications, different populations are compared using data that are sampled in a biased manner. Under sampling biases, standard methods that estimate the difference between the population means yield unreliable inferences. Here we develop an inference method that is resilient to sampling biases and is able to control the false positive errors under moderate bias levels in contrast to the standard approach. We demonstrate the method using synthetic and real biomarker data.

Acute kidney injury (AKI) commonly occurs in hospitalized patients and can lead to serious medical complications. In order to optimally predict AKI before it develops at any time during a hospital stay, we present a novel framework in which AKI is continually predicted automatically from EHR data over the entire hospital stay instead of at only one particular time. The continual model predicts AKI every time a patients AKI-relevant variable changes in the EHR. Thus the model is not only independent of a particular time for making predictions, but it can also leverage the latest values of all the AKI-relevant patient variables for making predictions. Using data of 44,691 hospital stays of duration longer than 24 hours we evaluated our continual prediction model and compared it with the traditional one-time prediction models. Excluding hospitals stays in which AKI occurred within 24 hours from admission, the one-time prediction model predicting at 24 hours from admission obtained area under ROC curve (AUC) of 0.653 while the continual prediction model obtained AUC of 0.724. The one-time prediction model that predicts at 24 hours obviously cannot predict AKI incidences that occur within 24 hours of admission which when included in the evaluation reduced its AUC to 0.57. In comparison, the continual prediction model had AUC of 0.709. The continual prediction model also did better than all other one-time prediction models predicting at other fixed times. By being able to take into account the latest values of AKI-relevant patient variables and by not being limited to a particular time of prediction, the continual prediction model out-performed one-time prediction models in predicting AKI.

We consider the task of localizing targets of interest in a hyperspectral (HS) image based on their spectral signature(s), by posing the problem as two distinct convex demixing task(s). With applications ranging from remote sensing to surveillance, this task of target detection leverages the fact that each material/object possesses its own characteristic spectral response, depending upon its composition. However, since $\textit{signatures}$ of different materials are often correlated, matched filtering-based approaches may not be apply here. To this end, we model a HS image as a superposition of a low-rank component and a dictionary sparse component, wherein the dictionary consists of the $\textit{a priori}$ known characteristic spectral responses of the target we wish to localize, and develop techniques for two different sparsity structures, resulting from different model assumptions. We also present the corresponding recovery guarantees, leveraging our recent theoretical results from a companion paper. Finally, we analyze the performance of the proposed approach via experimental evaluations on real HS datasets for a classification task, and compare its performance with related techniques.

Current learning machines have successfully solved hard application problems, reaching high accuracy and displaying seemingly "intelligent" behavior. Here we apply recent techniques for explaining decisions of state-of-the-art learning machines and analyze various tasks from computer vision and arcade games. This showcases a spectrum of problem-solving behaviors ranging from naive and short-sighted, to well-informed and strategic. We observe that standard performance evaluation metrics can be oblivious to distinguishing these diverse problem solving behaviors. Furthermore, we propose our semi-automated Spectral Relevance Analysis that provides a practically effective way of characterizing and validating the behavior of nonlinear learning machines. This helps to assess whether a learned model indeed delivers reliably for the problem that it was conceived for. Furthermore, our work intends to add a voice of caution to the ongoing excitement about machine intelligence and pledges to evaluate and judge some of these recent successes in a more nuanced manner.

The massive progress of machine learning has seen its application over a variety of domains in the past decade. But how do we develop a systematic, scalable and modular strategy to validate machine-learning systems? We present, to the best of our knowledge, the first approach, which provides a systematic test framework for machine-learning systems that accepts grammar-based inputs. Our OGMA approach automatically discovers erroneous behaviours in classifiers and leverages these erroneous behaviours to improve the respective models. OGMA leverages inherent robustness properties present in any well trained machine-learning model to direct test generation and thus, implementing a scalable test generation methodology. To evaluate our OGMA approach, we have tested it on three real world natural language processing (NLP) classifiers. We have found thousands of erroneous behaviours in these systems. We also compare OGMA with a random test generation approach and observe that OGMA is more effective than such random test generation by up to 489%.

Localization in challenging, natural environments such as forests or woodlands is an important capability for many applications from guiding a robot navigating along a forest trail to monitoring vegetation growth with handheld sensors. In this work we explore laser-based localization in both urban and natural environments, which is suitable for online applications. We propose a deep learning approach capable of learning meaningful descriptors directly from 3D point clouds by comparing triplets (anchor, positive and negative examples). The approach learns a feature space representation for a set of segmented point clouds that are matched between a current and previous observations. Our learning method is tailored towards loop closure detection resulting in a small model which can be deployed using only a CPU. The proposed learning method would allow the full pipeline to run on robots with limited computational payload such as drones, quadrupeds or UGVs.

Fairness is increasingly recognized as a critical component of machine learning systems. However, it is the underlying data on which these systems are trained that often reflect discrimination, suggesting a database repair problem. Existing treatments of fairness rely on statistical correlations that can be fooled by statistical anomalies, such as Simpson's paradox. Proposals for causality-based definitions of fairness can correctly model some of these situations, but they require specification of the underlying causal models. In this paper, we formalize the situation as a database repair problem, proving sufficient conditions for fair classifiers in terms of admissible variables as opposed to a complete causal model. We show that these conditions correctly capture subtle fairness violations. We then use these conditions as the basis for database repair algorithms that provide provable fairness guarantees about classifiers trained on their training labels. We evaluate our algorithms on real data, demonstrating improvement over the state of the art on multiple fairness metrics proposed in the literature while retaining high utility.

Hypothesis Testing, as such an important statistical technique applied widely in A/B testing for various business cases, has been relatively confusing to many people at the same time. This article aims to summarize the concept of a few key elements of hypothesis testing as well as how they impact the test results. The story starts from hypothesis. When we want to know any characteristics about a population like the form of distribution, the parameter of interest(mean, variance etc.), we make an assumption about it, which is called the hypothesis of population. Then we pull samples from population, and test whether the sample results make sense given the assumption. For example, your manager somehow knew that the mean of the click-through-rate per user from company's website across the user base is 0.06(mean of CTR of population), while you doubt that and believe the CTR should be higher.

In the syntax below, we are asking Python to import numpy and pandas package. The'as' is used to alias package name.