Search

Tabular datasets with low-sample-size or many variables are prevalent in biomedicine. Practitioners in this domain prefer linear or tree-based models over neural networks since the latter are harder to interpret and tend to overfit when applied to tabular datasets. To address these neural networks' shortcomings, we propose an intrinsically interpretable network for heterogeneous biomedical data. We design a locally sparse neural network where the local sparsity is learned to identify the subset of most relevant features for each sample. This sample-specific sparsity is predicted via a \textit{gating} network, which is trained in tandem with the \textit{prediction} network. By forcing the model to select a subset of the most informative features for each sample, we reduce model overfitting in low-sample-size data and obtain an interpretable model. We demonstrate that our method outperforms state-of-the-art models when applied to synthetic or real-world biomedical datasets using extensive experiments. Furthermore, the proposed framework dramatically outperforms existing schemes when evaluating its interpretability capabilities. Finally, we demonstrate the applicability of our model to two important biomedical tasks: survival analysis and marker gene identification.

We study the problem of exact support recovery based on noisy observations and present Refined Least Squares (RLS). Given a set of noisy measurement $$ \myvec{y} = \myvec{X}\myvec{\theta}^* + \myvec{\omega},$$ and $\myvec{X} \in \mathbb{R}^{N \times D}$ which is a (known) Gaussian matrix and $\myvec{\omega} \in \mathbb{R}^N$ is an (unknown) Gaussian noise vector, our goal is to recover the support of the (unknown) sparse vector $\myvec{\theta}^* \in \left\{-1,0,1\right\}^D$. To recover the support of the $\myvec{\theta}^*$ we use an average of multiple least squares solutions, each computed based on a subset of the full set of equations. The support is estimated by identifying the most significant coefficients of the average least squares solution. We demonstrate that in a wide variety of settings our method outperforms state-of-the-art support recovery algorithms.

Sleep disorders can be a negative factor both for learning as for the mental and physical development of adolescents. It has been shown that, in many populations, adolescents tend to have a poor sleep quality, and a very late chronotype. Furthermore, these features peak at adolescence, in the sense that adults tend to sleep better and have an earlier chronotype. But what happens when we consider adolescents in a population where already adults have poor sleep quality and a very late chronotype? We have conducted two non-clinical studies in the city of Bariloche, Argentina aimed at measuring sleep quality, chronotype, and social jet lag, using the Pittsburgh and Munich questionnaires. These were administered individually to groups of high school students, as well as to smaller samples of adults and preadolescents, in order to study differences between adolescents and these groups. The results show that in this population sleep quality is much poorer than in most other healthy populations recorded elsewhere. Furthermore, sleep quality is consistently worse for adolescents than for the other groups. The difference with adults seems to be due mainly to increased daytime sleepiness and sleep latency, whereas the difference with preadolescents seems to be due mainly to shorter sleep duration. We also found that the chronotypes of all the groups are very late, with a peak at an age between 18 and 24 ys. Social jet lag and sleep onset latency are also large, and they peak at adolescence, which suggests that they might be closely related to the large prevalence of poor sleep quality that we find in adolescents., Comment: 19 pages, 4 tables, 10 figures

In contrast to human vision, common recognition algorithms often fail on partially occluded images. We propose characterizing, empirically, the algorithmic limits by finding a minimal recognizable patch (MRP) that is by itself sufficient to recognize the image. A specialized deep network allows us to find the most informative patches of a given size, and serves as an experimental tool. A human vision study recently characterized related (but different) minimally recognizable configurations (MIRCs) [1], for which we specify computational analogues (denoted cMIRCs). The drop in human decision accuracy associated with size reduction of these MIRCs is substantial and sharp. Interestingly, such sharp reductions were also found for the computational versions we specified., Comment: Submitted to ICPR2020

Canonical Correlation Analysis (CCA) models are powerful for studying the associations between two sets of variables. The canonically correlated representations, termed \textit{canonical variates} are widely used in unsupervised learning to analyze unlabeled multi-modal registered datasets. Despite their success, CCA models may break (or overfit) if the number of variables in either of the modalities exceeds the number of samples. Moreover, often a significant fraction of the variables measures modality-specific information, and thus removing them is beneficial for identifying the \textit{canonically correlated variates}. Here, we propose $\ell_0$-CCA, a method for learning correlated representations based on sparse subsets of variables from two observed modalities. Sparsity is obtained by multiplying the input variables by stochastic gates, whose parameters are learned together with the CCA weights via an $\ell_0$-regularized correlation loss. We further propose $\ell_0$-Deep CCA for solving the problem of non-linear sparse CCA by modeling the correlated representations using deep nets. We demonstrate the efficacy of the method using several synthetic and real examples. Most notably, by gating nuisance input variables, our approach improves the extracted representations compared to other linear, non-linear and sparse CCA-based models.

A low-dimensional dynamical system is observed in an experiment as a high-dimensional signal; for example, a video of a chaotic pendulums system. Assuming that we know the dynamical model up to some unknown parameters, can we estimate the underlying system's parameters by measuring its time-evolution only once? The key information for performing this estimation lies in the temporal inter-dependencies between the signal and the model. We propose a kernel-based score to compare these dependencies. Our score generalizes a maximum likelihood estimator for a linear model to a general nonlinear setting in an unknown feature space. We estimate the system's underlying parameters by maximizing the proposed score. We demonstrate the accuracy and efficiency of the method using two chaotic dynamical systems - the double pendulum and the Lorenz '63 model.

Scientific observations may consist of a large number of variables (features). Identifying a subset of meaningful features is often ignored in unsupervised learning, despite its potential for unraveling clear patterns hidden in the ambient space. In this paper, we present a method for unsupervised feature selection, and we demonstrate its use for the task of clustering. We propose a differentiable loss function that combines the Laplacian score, which favors low-frequency features, with a gating mechanism for feature selection. We improve the Laplacian score, by replacing it with a gated variant computed on a subset of features. This subset is obtained using a continuous approximation of Bernoulli variables whose parameters are trained to gate the full feature space. We mathematically motivate the proposed approach and demonstrate that in the high noise regime, it is crucial to compute the Laplacian on the gated inputs, rather than on the full feature set. Experimental demonstration of the efficacy of the proposed approach and its advantage over current baselines is provided using several real-world examples.

We introduce a new deep learning method for point cloud comparison. Our approach, named Deep Point Cloud Distance (DPDist), measures the distance between the points in one cloud and the estimated surface from which the other point cloud is sampled. The surface is estimated locally and efficiently using the 3D modified Fisher vector representation. The local representation reduces the complexity of the surface, enabling efficient and effective learning, which generalizes well between object categories. We test the proposed distance in challenging tasks, such as similar object comparison and registration, and show that it provides significant improvements over commonly used distances such as Chamfer distance, Earth mover's distance, and others.

We propose a deep-learning based method for obtaining standardized data coordinates from scientific measurements.Data observations are modeled as samples from an unknown, non-linear deformation of an underlying Riemannian manifold, which is parametrized by a few normalized latent variables. By leveraging a repeated measurement sampling strategy, we present a method for learning an embedding in $\mathbb{R}^d$ that is isometric to the latent variables of the manifold. These data coordinates, being invariant under smooth changes of variables, enable matching between different instrumental observations of the same phenomenon. Our embedding is obtained using a LOcal Conformal Autoencoder (LOCA), an algorithm that constructs an embedding to rectify deformations by using a local z-scoring procedure while preserving relevant geometric information. We demonstrate the isometric embedding properties of LOCA on various model settings and observe that it exhibits promising interpolation and extrapolation capabilities. Finally, we apply LOCA to single-site Wi-Fi localization data, and to $3$-dimensional curved surface estimation based on a $2$-dimensional projection.

We study the problem of exact support recovery: given an (unknown) vector $\theta \in \left\{-1,0,1\right\}^D$, we are given access to the noisy measurement $$ y = X\theta + \omega,$$ where $X \in \mathbb{R}^{N \times D}$ is a (known) Gaussian matrix and the noise $\omega \in \mathbb{R}^N$ is an (unknown) Gaussian vector. How small we can choose $N$ and still reliably recover the support of $\theta$? We present RAWLS (Randomly Aggregated UnWeighted Least Squares Support Recovery): the main idea is to take random subsets of the $N$ equations, perform a least squares recovery over this reduced bit of information and then average over many random subsets. We show that the proposed procedure can provably recover an approximation of $\theta$ and demonstrate its use in support recovery through numerical examples.

word2vec due to Mikolov \textit{et al.} (2013) is a word embedding method that is widely used in natural language processing. Despite its great success and frequent use, theoretical justification is still lacking. The main contribution of our paper is to propose a rigorous analysis of the highly nonlinear functional of word2vec. Our results suggest that word2vec may be primarily driven by an underlying spectral method. This insight may open the door to obtaining provable guarantees for word2vec. We support these findings by numerical simulations. One fascinating open question is whether the nonlinear properties of word2vec that are not captured by the spectral method are beneficial and, if so, by what mechanism.

Organizations are starting to realize of the combined power of data and data-driven algorithmic models to gain insights, situational awareness, and advance their mission. A common challenge to gaining insights is connecting inherently different datasets. These datasets (e.g. geocoded features, video streams, raw text, social network data, etc.) per separate they provide very narrow answers; however collectively they can provide new capabilities. In this work, we present a data fusion framework for accelerating solutions for Processing, Exploitation, and Dissemination (PED). Our platform is a collection of services that extract information from several data sources (per separate) by leveraging deep learning and other means of processing. This information is fused by a set of analytical engines that perform data correlations, searches, and other modeling operations to combine information from the disparate data sources. As a result, events of interest are detected, geolocated, logged, and presented into a common operating picture. This common operating picture allows the user to visualize in real time all the data sources, per separate and their collective cooperation. In addition, forensic activities have been implemented and made available through the framework. Users can review archived results and compare them to the most recent snapshot of the operational environment. In our first iteration we have focused on visual data (FMV, WAMI, CCTV/PTZ-Cameras, open source video, etc.) and AIS data streams (satellite and terrestrial sources). As a proof-of-concept, in our experiments we show how FMV detections can be combined with vessel tracking signals from AIS sources to confirm identity, tip-and-cue aerial reconnaissance, and monitor vessel activity in an area.

With the proliferation of imaging sensors, the volume of multi-modal imagery far exceeds the ability of human analysts to adequately consume and exploit it. Full motion video (FMV) possesses the extra challenge of containing large amounts of redundant temporal data. We aim to address the needs of human analysts to consume and exploit data given aerial FMV. We have investigated and designed a system capable of detecting events and activities of interest that deviate from the baseline patterns of observation given FMV feeds. We have divided the problem into three tasks: (1) Context awareness, (2) object cataloging, and (3) event detection. The goal of context awareness is to constraint the problem of visual search and detection in video data. A custom image classifier categorizes the scene with one or multiple labels to identify the operating context and environment. This step helps reducing the semantic search space of downstream tasks in order to increase their accuracy. The second step is object cataloging, where an ensemble of object detectors locates and labels any known objects found in the scene (people, vehicles, boats, planes, buildings, etc.). Finally, context information and detections are sent to the event detection engine to monitor for certain behaviors. A series of analytics monitor the scene by tracking object counts, and object interactions. If these object interactions are not declared to be commonly observed in the current scene, the system will report, geolocate, and log the event. Events of interest include identifying a gathering of people as a meeting and/or a crowd, alerting when there are boats on a beach unloading cargo, increased count of people entering a building, people getting in and/or out of vehicles of interest, etc. We have applied our methods on data from different sensors at different resolutions in a variety of geographical areas.

Current successful approaches for generic (non-semantic) segmentation rely mostly on edge detection and have leveraged the strengths of deep learning mainly by improving the edge detection stage in the algorithmic pipeline. This is in contrast to semantic and instance segmentation, where DNNs are applied directly to generate pixel-wise segment representations. We propose a new method for learning a pixel-wise representation that reflects segment relatedness. This representation is combined with an edge map to yield a new segmentation algorithm. We show that the representations themselves achieve state-of-the-art segment similarity scores. Moreover, the proposed combined segmentation algorithm provides results that are either state of the art or improve upon it, for most quality measures., Comment: CVPR 2020. arXiv admin note: substantial text overlap with arXiv:1909.11735

Deep learning approaches to generic (non-semantic) segmentation have so far been indirect and relied on edge detection. This is in contrast to semantic segmentation, where DNNs are applied directly. We propose an alternative approach called Deep Generic Segmentation (DGS) and try to follow the path used for semantic segmentation. Our main contribution is a new method for learning a pixel-wise representation that reflects segment relatedness. This representation is combined with a CRF to yield the segmentation algorithm. We show that we are able to learn meaningful representations that improve segmentation quality and that the representations themselves achieve state-of-the-art segment similarity scores. The segmentation results are competitive and promising.

Humans possess an intricate and powerful visual system in order to perceive and understand the environing world. Human perception can effortlessly detect and correctly group features in visual data and can even interpret random-dot videos induced by imaging natural dynamic scenes with highly noisy sensors such as ultrasound imaging. Remarkably, this happens even if perception completely fails when the same information is presented frame by frame rather than in a video sequence. We study this property of surprising dynamic perception with the first goal of proposing a new detection and spatio-temporal grouping algorithm for such signals when, per frame, the information on objects is both random and sparse and embedded in random noise. The algorithm is based on the succession of temporal integration and spatial statistical tests of unlikeliness, the a contrario framework. The algorithm not only manages to handle such signals but the striking similarity in its performance to the perception by human observers, as witnessed by a series of psychophysical experiments on image and video data, leads us to see in it a simple computational Gestalt model of human perception with only two parameters: the time integration and the visual angle for candidate shapes to be detected., Comment: Technical report ; Resubmission: Corrected typos. Rephrased some sentences

Variational autoencoders (VAEs) and generative adversarial networks (GANs) enjoy an intuitive connection to manifold learning: in training the decoder/generator is optimized to approximate a homeomorphism between the data distribution and the sampling space. This is a construction that strives to define the data manifold. A major obstacle to VAEs and GANs, however, is choosing a suitable prior that matches the data topology. Well-known consequences of poorly picked priors are posterior and mode collapse. To our knowledge, no existing method sidesteps this user choice. Conversely, $\textit{diffusion maps}$ automatically infer the data topology and enjoy a rigorous connection to manifold learning, but do not scale easily or provide the inverse homeomorphism (i.e. decoder/generator). We propose a method that combines these approaches into a generative model that inherits the asymptotic guarantees of $\textit{diffusion maps}$ while preserving the scalability of deep models. We prove approximation theoretic results for the dimension dependence of our proposed method. Finally, we demonstrate the effectiveness of our method with various real and synthetic datasets., Comment: 24 pages, 9 figures, 1 table; accepted to ECCV 2020

Detection and segmentation of objects in overheard imagery is a challenging task. The variable density, random orientation, small size, and instance-to-instance heterogeneity of objects in overhead imagery calls for approaches distinct from existing models designed for natural scene datasets. Though new overhead imagery datasets are being developed, they almost universally comprise a single view taken from directly overhead ("at nadir"), failing to address a critical variable: look angle. By contrast, views vary in real-world overhead imagery, particularly in dynamic scenarios such as natural disasters where first looks are often over 40 degrees off-nadir. This represents an important challenge to computer vision methods, as changing view angle adds distortions, alters resolution, and changes lighting. At present, the impact of these perturbations for algorithmic detection and segmentation of objects is untested. To address this problem, we present an open source Multi-View Overhead Imagery dataset, termed SpaceNet MVOI, with 27 unique looks from a broad range of viewing angles (-32.5 degrees to 54.0 degrees). Each of these images cover the same 665 square km geographic extent and are annotated with 126,747 building footprint labels, enabling direct assessment of the impact of viewpoint perturbation on model performance. We benchmark multiple leading segmentation and object detection models on: (1) building detection, (2) generalization to unseen viewing angles and resolutions, and (3) sensitivity of building footprint extraction to changes in resolution. We find that state of the art segmentation and object detection models struggle to identify buildings in off-nadir imagery and generalize poorly to unseen views, presenting an important benchmark to explore the broadly relevant challenge of detecting small, heterogeneous target objects in visually dynamic contexts., Comment: Accepted into IEEE International Conference on Computer Vision (ICCV) 2019

In this paper, we propose a normal estimation method for unstructured 3D point clouds. This method, called Nesti-Net, builds on a new local point cloud representation which consists of multi-scale point statistics (MuPS), estimated on a local coarse Gaussian grid. This representation is a suitable input to a CNN architecture. The normals are estimated using a mixture-of-experts (MoE) architecture, which relies on a data-driven approach for selecting the optimal scale around each point and encourages sub-network specialization. Interesting insights into the network's resource distribution are provided. The scale prediction significantly improves robustness to different noise levels, point density variations and different levels of detail. We achieve state-of-the-art results on a benchmark synthetic dataset and present qualitative results on real scanned scenes., Comment: Code will be available after publication. Figure quality reduced to fit size requirement. Higher quality images will be available in the final paper

Feature selection problems have been extensively studied for linear estimation, for instance, Lasso, but less emphasis has been placed on feature selection for non-linear functions. In this study, we propose a method for feature selection in high-dimensional non-linear function estimation problems. The new procedure is based on minimizing the $\ell_0$ norm of the vector of indicator variables that represent if a feature is selected or not. Our approach relies on the continuous relaxation of Bernoulli distributions, which allows our model to learn the parameters of the approximate Bernoulli distributions via gradient descent. This general framework simultaneously minimizes a loss function while selecting relevant features. Furthermore, we provide an information-theoretic justification of incorporating Bernoulli distribution into our approach and demonstrate the potential of the approach on synthetic and real-life applications., Comment: Published in ICML 2020

Foundational mapping remains a challenge in many parts of the world, particularly in dynamic scenarios such as natural disasters when timely updates are critical. Updating maps is currently a highly manual process requiring a large number of human labelers to either create features or rigorously validate automated outputs. We propose that the frequent revisits of earth imaging satellite constellations may accelerate existing efforts to quickly update foundational maps when combined with advanced machine learning techniques. Accordingly, the SpaceNet partners (CosmiQ Works, Radiant Solutions, and NVIDIA), released a large corpus of labeled satellite imagery on Amazon Web Services (AWS) called SpaceNet. The SpaceNet partners also launched a series of public prize competitions to encourage improvement of remote sensing machine learning algorithms. The first two of these competitions focused on automated building footprint extraction, and the most recent challenge focused on road network extraction. In this paper we discuss the SpaceNet imagery, labels, evaluation metrics, prize challenge results to date, and future plans for the SpaceNet challenge series., Comment: 10 pages, 5 figures, 2 tables, 5 appendices

We present the DeepGlobe 2018 Satellite Image Understanding Challenge, which includes three public competitions for segmentation, detection, and classification tasks on satellite images. Similar to other challenges in computer vision domain such as DAVIS and COCO, DeepGlobe proposes three datasets and corresponding evaluation methodologies, coherently bundled in three competitions with a dedicated workshop co-located with CVPR 2018. We observed that satellite imagery is a rich and structured source of information, yet it is less investigated than everyday images by computer vision researchers. However, bridging modern computer vision with remote sensing data analysis could have critical impact to the way we understand our environment and lead to major breakthroughs in global urban planning or climate change research. Keeping such bridging objective in mind, DeepGlobe aims to bring together researchers from different domains to raise awareness of remote sensing in the computer vision community and vice-versa. We aim to improve and evaluate state-of-the-art satellite image understanding approaches, which can hopefully serve as reference benchmarks for future research in the same topic. In this paper, we analyze characteristics of each dataset, define the evaluation criteria of the competitions, and provide baselines for each task., Comment: Dataset description for DeepGlobe 2018 Challenge at CVPR 2018

Many generative models attempt to replicate the density of their input data. However, this approach is often undesirable, since data density is highly affected by sampling biases, noise, and artifacts. We propose a method called SUGAR (Synthesis Using Geometrically Aligned Random-walks) that uses a diffusion process to learn a manifold geometry from the data. Then, it generates new points evenly along the manifold by pulling randomly generated points into its intrinsic structure using a diffusion kernel. SUGAR equalizes the density along the manifold by selectively generating points in sparse areas of the manifold. We demonstrate how the approach corrects sampling biases and artifacts, while also revealing intrinsic patterns (e.g. progression) and relations in the data. The method is applicable for correcting missing data, finding hypothetical data points, and learning relationships between data features.

The point cloud is gaining prominence as a method for representing 3D shapes, but its irregular format poses a challenge for deep learning methods. The common solution of transforming the data into a 3D voxel grid introduces its own challenges, mainly large memory size. In this paper we propose a novel 3D point cloud representation called 3D Modified Fisher Vectors (3DmFV). Our representation is hybrid as it combines the discrete structure of a grid with continuous generalization of Fisher vectors, in a compact and computationally efficient way. Using the grid enables us to design a new CNN architecture for point cloud classification and part segmentation. In a series of experiments we demonstrate competitive performance or even better than state-of-the-art on challenging benchmark datasets.

Using image context is an effective approach for improving object detection. Previously proposed methods used contextual cues that rely on semantic or spatial information. In this work, we explore a different kind of contextual information: inner-scene similarity. We present the CISS (Context by Inner Scene Similarity) algorithm, which is based on the observation that two visually similar sub-image patches are likely to share semantic identities, especially when both appear in the same image. CISS uses base-scores provided by a base detector and performs as a post-detection stage. For each candidate sub-image (denoted anchor), the CISS algorithm finds a few similar sub-images (denoted supporters), and, using them, calculates a new enhanced score for the anchor. This is done by utilizing the base-scores of the supporters and a pre-trained dependency model. The new scores are modeled as a linear function of the base scores of the anchor and the supporters and is estimated using a minimum mean square error optimization. This approach results in: (a) improved detection of partly occluded objects (when there are similar non-occluded objects in the scene), and (b) fewer false alarms (when the base detector mistakenly classifies a background patch as an object). This work relates to Duncan and Humphreys' "similarity theory," a psychophysical study. which suggested that the human visual system perceptually groups similar image regions and that the classification of one region is affected by the estimated identity of the other. Experimental results demonstrate the enhancement of a base detector's scores on the PASCAL VOC dataset.

Kernel methods play a critical role in many machine learning algorithms. They are useful in manifold learning, classification, clustering and other data analysis tasks. Setting the kernel's scale parameter, also referred to as the kernel's bandwidth, highly affects the performance of the task in hand. We propose to set a scale parameter that is tailored to one of two types of tasks: classification and manifold learning. For manifold learning, we seek a scale which is best at capturing the manifold's intrinsic dimension. For classification, we propose three methods for estimating the scale, which optimize the classification results in different senses. The proposed frameworks are simulated on artificial and on real datasets. The results show a high correlation between optimal classification rates and the estimated scales. Finally, we demonstrate the approach on a seismic event classification task.

The problem of learning from seismic recordings has been studied for years. There is a growing interest in developing automatic mechanisms for identifying the properties of a seismic event. One main motivation is the ability have a reliable identification of man-made explosions. The availability of multiple high-dimensional observations has increased the use of machine learning techniques in a variety of fields. In this work, we propose to use a kernel-fusion based dimensionality reduction framework for generating meaningful seismic representations from raw data. The proposed method is tested on 2023 events that were recorded in Israel and in Jordan. The method achieves promising results in classification of event type as well as in estimating the location of the event. The proposed fusion and dimensionality reduction tools may be applied to other types of geophysical data.

Over-segmentation, or super-pixel generation, is a common preliminary stage for many computer vision applications. New acquisition technologies enable the capturing of 3D point clouds that contain color and geometrical information. This 3D information introduces a new conceptual change that can be utilized to improve the results of over-segmentation, which uses mainly color information, and to generate clusters of points we call super-points. We consider a variety of possible 3D extensions of the Local Variation (LV) graph based over-segmentation algorithms, and compare them thoroughly. We consider different alternatives for constructing the connectivity graph, for assigning the edge weights, and for defining the merge criterion, which must now account for the geometric information and not only color. Following this evaluation, we derive a new generic algorithm for over-segmentation of 3D point clouds. We call this new algorithm Point Cloud Local Variation (PCLV). The advantages of the new over-segmentation algorithm are demonstrated on both outdoor and cluttered indoor scenes. Performance analysis of the proposed approach compared to state-of-the-art 2D and 3D over-segmentation algorithms shows significant improvement according to the common performance measures.