Learning from imbalanced data is among the most challenging areas in contemporary machine learning. This becomes even more difficult when considered the context of big data that calls for dedicated architectures capable of high-performance processing. Apache Spark is a highly efficient and popular architecture, but it poses specific challenges for algorithms to be implemented for it. While oversampling algorithms are an effective way for handling class imbalance, they have not been designed for distributed environments. In this paper, we propose a holistic look on oversampling algorithms for imbalanced big data. We discuss the taxonomy of oversampling algorithms and their mechanisms used to handle skewed class distributions. We introduce a Spark library with 14 state-of-the-art oversampling algorithms implemented and evaluate their efficacy via extensive experimental study. Using binary and multi-class massive data sets, we analyze the effectiveness of oversampling algorithms and their relationships with different types of classifiers. We evaluate the trade-off between accuracy and time complexity of oversampling algorithms, as well as their scalability when increasing the size of data. This allows us to gain insight into the usefulness of specific components of oversampling algorithms for big data, as well as formulate guidelines and recommendations for designing future resampling approaches for massive imbalanced data. Our library can be downloaded from https://github.com/fsleeman/spark-class-balancing.git.
Liquid Argon Time Projection Chambers (LArTPC) are particle imaging detectors recording 2D or 3D images of trajectories of charged particles. Identifying points of interest in these images, namely the initial and terminal points of track-like particle trajectories such as muons and protons, and the initial points of electromagnetic shower-like particle trajectories such as electrons and gamma rays, is a crucial step of identifying and analyzing these particles and impacts the inference of physics signals such as neutrino interaction. The Point Proposal Network is designed to discover these specific points of interest. The algorithm predicts with a sub-voxel precision their spatial location, and also determines the category of the identified points of interest. Using as a benchmark the PILArNet public LArTPC data sample in which the voxel resolution is 3mm/voxel, our algorithm successfully predicted 96.8% and 97.8% of 3D points within a distance of 3 and 10~voxels from the provided true point locations respectively. For the predicted 3D points within 3 voxels of the closest true point locations, the median distance is found to be 0.25 voxels, achieving the sub-voxel level precision. In addition, we report our analysis of the mistakes where our algorithm prediction differs from the provided true point positions by more than 10~voxels. Among 50 mistakes visually scanned, 25 were due to the definition of true position location, 15 were legitimate mistakes where a physicist cannot visually disagree with the algorithm's prediction, and 10 were genuine mistakes that we wish to improve in the future. Further, using these predicted points, we demonstrate a simple algorithm to cluster 3D voxels into individual track-like particle trajectories with a clustering efficiency, purity, and Adjusted Rand Index of 96%, 93%, and 91% respectively.
This paper presents the details of the SRIB-LEAP submission to the ConferencingSpeech challenge 2021. The challenge involved the task of multi-channel speech enhancement to improve the quality of far field speech from microphone arrays in a video conferencing room. We propose a two stage method involving a beamformer followed by single channel enhancement. For the beamformer, we incorporated self-attention mechanism as inter-channel processing layer in the filter-and-sum network (FaSNet), an end-to-end time-domain beamforming system. The single channel speech enhancement is done in log spectral domain using convolution neural network (CNN)-long short term memory (LSTM) based architecture. We achieved improvements in objective quality metrics - perceptual evaluation of speech quality (PESQ) of 0.5 on the noisy data. On subjective quality evaluation, the proposed approach improved the mean opinion score (MOS) by an absolute measure of 0.9 over the noisy audio.
We investigate the efficiency of two very different spoken term detection approaches for transcription when the available data is insufficient to train a robust ASR system. This work is grounded in very low-resource language documentation scenario where only few minutes of recording have been transcribed for a given language so far.Experiments on two oral languages show that a pretrained universal phone recognizer, fine-tuned with only a few minutes of target language speech, can be used for spoken term detection with a better overall performance than a dynamic time warping approach. In addition, we show that representing phoneme recognition ambiguity in a graph structure can further boost the recall while maintaining high precision in the low resource spoken term detection task.
We present a data-driven optimization framework for redesigning police patrol zones in an urban environment. The objectives are to rebalance police workload among geographical areas and to reduce response time to emergency calls. We develop a stochastic model for police emergency response by integrating multiple data sources, including police incidents reports, demographic surveys, and traffic data. Using this stochastic model, we optimize zone redesign plans using mixed-integer linear programming. Our proposed design was implemented by the Atlanta Police Department in March 2019. By analyzing data before and after the zone redesign, we show that the new design has reduced the response time to high priority 911 calls by 5.8\% and the imbalance of police workload among different zones by 43\%.
Although Sinkhorn divergences are now routinely used in data sciences to compare probability distributions, the computational effort required to compute them remains expensive, growing in general quadratically in the size $n$ of the support of these distributions. Indeed, solving optimal transport (OT) with an entropic regularization requires computing a $n\times n$ kernel matrix (the neg-exponential of a $n\times n$ pairwise ground cost matrix) that is repeatedly applied to a vector. We propose to use instead ground costs of the form $c(x,y)=-\log\dotp{\varphi(x)}{\varphi(y)}$ where $\varphi$ is a map from the ground space onto the positive orthant $\RR^r_+$, with $r\ll n$. This choice yields, equivalently, a kernel $k(x,y)=\dotp{\varphi(x)}{\varphi(y)}$, and ensures that the cost of Sinkhorn iterations scales as $O(nr)$. We show that usual cost functions can be approximated using this form. Additionaly, we take advantage of the fact that our approach yields approximation that remain fully differentiable with respect to input distributions, as opposed to previously proposed adaptive low-rank approximations of the kernel matrix, to train a faster variant of OT-GAN \cite{salimans2018improving}.
We present a workflow for clinical data analysis that relies on Bayesian Structure Learning (BSL), an unsupervised learning approach, robust to noise and biases, that allows to incorporate prior medical knowledge into the learning process and that provides explainable results in the form of a graph showing the causal connections among the analyzed features. The workflow consists in a multi-step approach that goes from identifying the main causes of patient's outcome through BSL, to the realization of a tool suitable for clinical practice, based on a Binary Decision Tree (BDT), to recognize patients at high-risk with information available already at hospital admission time. We evaluate our approach on a feature-rich COVID-19 dataset, showing that the proposed framework provides a schematic overview of the multi-factorial processes that jointly contribute to the outcome. We discuss how these computational findings are confirmed by current understanding of the COVID-19 pathogenesis. Further, our approach yields to a highly interpretable tool correctly predicting the outcome of 85% of subjects based exclusively on 3 features: age, a previous history of chronic obstructive pulmonary disease and the PaO2/FiO2 ratio at the time of arrival to the hospital. The inclusion of additional information from 4 routine blood tests (Creatinine, Glucose, pO2 and Sodium) increases predictive accuracy to 94.5%.
Recovering a planted vector $v$ in an $n$-dimensional random subspace of $\mathbb{R}^N$ is a generic task related to many problems in machine learning and statistics, such as dictionary learning, subspace recovery, and principal component analysis. In this work, we study computationally efficient estimation and detection of a planted vector $v$ whose $\ell_4$ norm differs from that of a Gaussian vector with the same $\ell_2$ norm. For instance, in the special case of an $N \rho$-sparse vector $v$ with Rademacher nonzero entries, our results include the following: (1) We give an improved analysis of (a slight variant of) the spectral method proposed by Hopkins, Schramm, Shi, and Steurer, showing that it approximately recovers $v$ with high probability in the regime $n \rho \ll \sqrt{N}$. In contrast, previous work required either $\rho \ll 1/\sqrt{n}$ or $n \sqrt{\rho} \lesssim \sqrt{N}$ for polynomial-time recovery. Our result subsumes both of these conditions (up to logarithmic factors) and also treats the dense case $\rho = 1$ which was not previously considered. (2) Akin to $\ell_\infty$ bounds for eigenvector perturbation, we establish an entrywise error bound for the spectral estimator via a leave-one-out analysis, from which it follows that thresholding recovers $v$ exactly. (3) We study the associated detection problem and show that in the regime $n \rho \gg \sqrt{N}$, any spectral method from a large class (and more generally, any low-degree polynomial of the input) fails to detect the planted vector. This establishes optimality of our upper bounds and offers evidence that no polynomial-time algorithm can succeed when $n \rho \gg \sqrt{N}$.
Tracking a time-varying indefinite number of objects in a video sequence over time remains a challenge despite recent advances in the field. Ignoring long-term temporal information, most existing approaches are not able to properly handle multi-object tracking challenges such as occlusion. To address these shortcomings, we present MO3TR: a truly end-to-end Transformer-based online multi-object tracking (MOT) framework that learns to handle occlusions, track initiation and termination without the need for an explicit data association module or any heuristics/post-processing. MO3TR encodes object interactions into long-term temporal embeddings using a combination of spatial and temporal Transformers, and recursively uses the information jointly with the input data to estimate the states of all tracked objects over time. The spatial attention mechanism enables our framework to learn implicit representations between all the objects and the objects to the measurements, while the temporal attention mechanism focuses on specific parts of past information, allowing our approach to resolve occlusions over multiple frames. Our experiments demonstrate the potential of this new approach, reaching new state-of-the-art results on multiple MOT metrics for two popular multi-object tracking benchmarks. Our code will be made publicly available.
Although deep learning has demonstrated astonishing performance in many applications, there are still concerns about its dependability. One desirable property of deep learning applications with societal impact is fairness (i.e., non-discrimination). Unfortunately, discrimination might be intrinsically embedded into the models due to the discrimination in the training data. As a countermeasure, fairness testing systemically identifies discriminatory samples, which can be used to retrain the model and improve the model's fairness. Existing fairness testing approaches however have two major limitations. Firstly, they only work well on traditional machine learning models and have poor performance (e.g., effectiveness and efficiency) on deep learning models. Secondly, they only work on simple structured (e.g., tabular) data and are not applicable for domains such as text. In this work, we bridge the gap by proposing a scalable and effective approach for systematically searching for discriminatory samples while extending existing fairness testing approaches to address a more challenging domain, i.e., text classification. Compared with state-of-the-art methods, our approach only employs lightweight procedures like gradient computation and clustering, which is significantly more scalable and effective. Experimental results show that on average, our approach explores the search space much more effectively (9.62 and 2.38 times more than the state-of-the-art methods respectively on tabular and text datasets) and generates much more discriminatory samples (24.95 and 2.68 times) within a same reasonable time. Moreover, the retrained models reduce discrimination by 57.2% and 60.2% respectively on average.