A survey on 3D object detection in real time for autonomous driving.

This survey reviews advances in 3D object detection approaches for autonomous driving. A brief introduction to 2D object detection is first discussed and drawbacks of the existing methodologies are identified for highly dynamic environments. Subsequently, this paper reviews the state-of-the-art 3D object detection techniques that utilizes monocular and stereo vision for reliable detection in urban settings. Based on depth inference basis, learning schemes, and internal representation, this work presents a method taxonomy of three classes: model-based and geometrically constrained approaches, end-to-end learning methodologies, and hybrid methods. There is highlighted segment for current trend of multi-view detectors as end-to-end methods due to their boosted robustness. Detectors from the last two kinds were specially selected to exploit the autonomous driving context in terms of geometry, scene content and instances distribution. To prove the effectiveness of each method, 3D object detection datasets for autonomous vehicles are described with their unique features, e. g., varying weather conditions, multi-modality, multi camera perspective and their respective metrics associated to different difficulty categories. In addition, we included multi-modal visual datasets, i. e., V2X that may tackle the problems of single-view occlusion. Finally, the current research trends in object detection are summarized, followed by a discussion on possible scope for future research in this domain.

Applying machine learning to predict viral assembly for adeno-associated virus capsid libraries.

Machine learning (ML) can aid in novel discoveries in the field of viral gene therapy. Specifically, big data gathered through next-generation sequencing (NGS) of complex capsid libraries is an especially prominent source of lost potential in data analysis and prediction. Furthermore, adeno-associated virus (AAV)-based capsid libraries are becoming increasingly popular as a tool to select candidates for gene therapy vectors. These higher complexity AAV capsid libraries have previously been created and selected in vivo; however, in silico analysis using ML computer algorithms may augment smarter and more robust libraries for selection. In this study, data of AAV capsid libraries gathered before and after viral assembly are used to train ML algorithms. We found that two ML computer algorithms, artificial neural networks (ANNs), and support vector machines (SVMs), can be trained to predict whether unknown capsid variants may assemble into viable virus-like structures. Using the most accurate models constructed, hypothetical mutation patterns in library construction were simulated to suggest the importance of N495, G546, and I554 in AAV2-derived capsids. Finally, two comparative libraries were generated using ML-derived data to biologically validate these findings and demonstrate the predictive power of ML in vector design.

An artificial delay based robust guidance strategy for an interceptor with input saturation.

This paper proposes a time-energy efficient, artificial time delay based robust guidance strategy with input saturation for a two-dimensional interceptor problem. A reference near optimal heading trajectory is generated by applying Differential Evolution (DE) to the interceptor problem. By following the reference heading trajectory, the robust control law guides the missile to intercept the target in a time-energy efficient manner, while tackling the disturbances and uncertainties that it might encounter. The near optimal trajectory is obtained offline, whereas the robust guidance strategy has been applied online which further increases the appeal of the proposed guidance scheme. Uniformly ultimately bounded (UUB) stability has been affirmed for the closed loop system employing Lyapunov's method. Also, the proposed guidance law has been tested through simulation on both non-maneuvering and maneuvering targets performing bank to bank as well as step maneuver in the presence of uncertainties.

Joint genome-wide prediction in several populations accounting for randomness of genotypes: A hierarchical Bayes approach. I: Multivariate Gaussian priors for marker effects and derivation of the joint probability mass function of genotypes.

It is important to consider heterogeneity of marker effects and allelic frequencies in across population genome-wide prediction studies. Moreover, all regression models used in genome-wide prediction overlook randomness of genotypes. In this study, a family of hierarchical Bayesian models to perform across population genome-wide prediction modeling genotypes as random variables and allowing population-specific effects for each marker was developed. Models shared a common structure and differed in the priors used and the assumption about residual variances (homogeneous or heterogeneous). Randomness of genotypes was accounted for by deriving the joint probability mass function of marker genotypes conditional on allelic frequencies and pedigree information. As a consequence, these models incorporated kinship and genotypic information that not only permitted to account for heterogeneity of allelic frequencies, but also to include individuals with missing genotypes at some or all loci without the need for previous imputation. This was possible because the non-observed fraction of the design matrix was treated as an unknown model parameter. For each model, a simpler version ignoring population structure, but still accounting for randomness of genotypes was proposed. Implementation of these models and computation of some criteria for model comparison were illustrated using two simulated datasets. Theoretical and computational issues along with possible applications, extensions and refinements were discussed. Some features of the models developed in this study make them promising for genome-wide prediction, the use of information contained in the probability distribution of genotypes is perhaps the most appealing. Further studies to assess the performance of the models proposed here and also to compare them with conventional models used in genome-wide prediction are needed.

Joint genome-wide prediction in several populations accounting for randomness of genotypes: A hierarchical Bayes approach. II: Multivariate spike and slab priors for marker effects and derivation of approximate Bayes and fractional Bayes factors for the complete family of models.

This study corresponds to the second part of a companion paper devoted to the development of Bayesian multiple regression models accounting for randomness of genotypes in across population genome-wide prediction. This family of models considers heterogeneous and correlated marker effects and allelic frequencies across populations, and has the ability of considering records from non-genotyped individuals and individuals with missing genotypes in any subset of loci without the need for previous imputation, taking into account uncertainty about imputed genotypes. This paper extends this family of models by considering multivariate spike and slab conditional priors for marker allele substitution effects and contains derivations of approximate Bayes factors and fractional Bayes factors to compare models from part I and those developed here with their null versions. These null versions correspond to simpler models ignoring heterogeneity of populations, but still accounting for randomness of genotypes. For each marker loci, the spike component of priors corresponded to point mass at 0 in RS, where S is the number of populations, and the slab component was a S-variate Gaussian distribution, independent conditional priors were assumed. For the Gaussian components, covariance matrices were assumed to be either the same for all markers or different for each marker. For null models, the priors were simply univariate versions of these finite mixture distributions. Approximate algebraic expressions for Bayes factors and fractional Bayes factors were found using the Laplace approximation. Using the simulated datasets described in part I, these models were implemented and compared with models derived in part I using measures of predictive performance based on squared Pearson correlations, Deviance Information Criterion, Bayes factors, and fractional Bayes factors. The extensions presented here enlarge our family of genome-wide prediction models making it more flexible in the sense that it now offers more modeling options.

Learning Precise Spike Train-to-Spike Train Transformations in Multilayer Feedforward Neuronal Networks.

We derive a synaptic weight update rule for learning temporally precise spike train-to-spike train transformations in multilayer feedforward networks of spiking neurons. The framework, aimed at seamlessly generalizing error backpropagation to the deterministic spiking neuron setting, is based strictly on spike timing and avoids invoking concepts pertaining to spike rates or probabilistic models of spiking. The derivation is founded on two innovations. First, an error functional is proposed that compares the spike train emitted by the output neuron of the network to the desired spike train by way of their putative impact on a virtual postsynaptic neuron. This formulation sidesteps the need for spike alignment and leads to closed-form solutions for all quantities of interest. Second, virtual assignment of weights to spikes rather than synapses enables a perturbation analysis of individual spike times and synaptic weights of the output, as well as all intermediate neurons in the network, which yields the gradients of the error functional with respect to the said entities. Learning proceeds via a gradient descent mechanism that leverages these quantities. Simulation experiments demonstrate the efficacy of the proposed learning framework. The experiments also highlight asymmetries between synapses on excitatory and inhibitory neurons.

Bayclone: Bayesian nonparametric inference of tumor subclones using NGS data.

In this paper, we present a novel feature allocation model to describe tumor heterogeneity (TH) using next-generation sequencing (NGS) data. Taking a Bayesian approach, we extend the Indian buffet process (IBP) to define a class of nonparametric models, the categorical IBP (cIBP). A cIBP takes categorical values to denote homozygous or heterozygous genotypes at each SNV. We define a subclone as a vector of these categorical values, each corresponding to an SNV. Instead of partitioning somatic mutations into non-overlapping clusters with similar cellular prevalences, we took a different approach using feature allocation. Importantly, we do not assume somatic mutations with similar cellular prevalence must be from the same subclone and allow overlapping mutations shared across subclones. We argue that this is closer to the underlying theory of phylogenetic clonal expansion, as somatic mutations occurred in parent subclones should be shared across the parent and child subclones. Bayesian inference yields posterior probabilities of the number, genotypes, and proportions of subclones in a tumor sample, thereby providing point estimates as well as variabilities of the estimates for each subclone. We report results on both simulated and real data. BayClone is available at http://health.bsd.uchicago.edu/yji/soft.html.

Connectomic constraints on computation in feedforward networks of spiking neurons.

Several efforts are currently underway to decipher the connectome or parts thereof in a variety of organisms. Ascertaining the detailed physiological properties of all the neurons in these connectomes, however, is out of the scope of such projects. It is therefore unclear to what extent knowledge of the connectome alone will advance a mechanistic understanding of computation occurring in these neural circuits, especially when the high-level function of the said circuit is unknown. We consider, here, the question of how the wiring diagram of neurons imposes constraints on what neural circuits can compute, when we cannot assume detailed information on the physiological response properties of the neurons. We call such constraints-that arise by virtue of the connectome-connectomic constraints on computation. For feedforward networks equipped with neurons that obey a deterministic spiking neuron model which satisfies a small number of properties, we ask if just by knowing the architecture of a network, we can rule out computations that it could be doing, no matter what response properties each of its neurons may have. We show results of this form, for certain classes of network architectures. On the other hand, we also prove that with the limited set of properties assumed for our model neurons, there are fundamental limits to the constraints imposed by network structure. Thus, our theory suggests that while connectomic constraints might restrict the computational ability of certain classes of network architectures, we may require more elaborate information on the properties of neurons in the network, before we can discern such results for other classes of networks.

Image denoising using the higher order singular value decomposition.

In this paper, we propose a very simple and elegant patch-based, machine learning technique for image denoising using the higher order singular value decomposition (HOSVD). The technique simply groups together similar patches from a noisy image (with similarity defined by a statistically motivated criterion) into a 3D stack, computes the HOSVD coefficients of this stack, manipulates these coefficients by hard thresholding, and inverts the HOSVD transform to produce the final filtered image. Our technique chooses all required parameters in a principled way, relating them to the noise model. We also discuss our motivation for adopting the HOSVD as an appropriate transform for image denoising. We experimentally demonstrate the excellent performance of the technique on grayscale as well as color images. On color images, our method produces state-of-the-art results, outperforming other color image denoising algorithms at moderately high noise levels. A criterion for optimal patch-size selection and noise variance estimation from the residual images (after denoising) is also presented.

Trainable Convolution Filters and Their Application to Face Recognition.

In this paper, we present a novel image classification system that is built around a core of trainable filter ensembles that we call Volterra kernel classifiers. Our system treats images as a collection of possibly overlapping patches and is composed of three components: (1) A scheme for a single patch classification that seeks a smooth, possibly nonlinear, functional mapping of the patches into a range space, where patches of the same class are close to one another, while patches from different classes are far apart-in the L_2 sense. This mapping is accomplished using trainable convolution filters (or Volterra kernels) where the convolution kernel can be of any shape or order. (2) Given a corpus of Volterra classifiers with various kernel orders and shapes for each patch, a boosting scheme for automatically selecting the best weighted combination of the classifiers to achieve higher per-patch classification rate. (3) A scheme for aggregating the classification information obtained for each patch via voting for the parent image classification. We demonstrate the effectiveness of the proposed technique using face recognition as an application area and provide extensive experiments on the Yale, CMU PIE, Extended Yale B, Multi-PIE, and MERL Dome benchmark face data sets. We call the Volterra kernel classifiers applied to face recognition Volterrafaces. We show that our technique, which falls into the broad class of embedding-based face image discrimination methods, consistently outperforms various state-of-the-art methods in the same category.

A finite-sample, distribution-free, probabilistic lower bound on mutual information.

For any memoryless communication channel with a binary-valued input and a one-dimensional real-valued output, we introduce a probabilistic lower bound on the mutual information given empirical observations on the channel. The bound is built on the Dvoretzky-Kiefer-Wolfowitz inequality and is distribution free. A quadratic time algorithm is described for computing the bound and its corresponding class-conditional distribution functions. We compare our approach to existing techniques and show the superiority of our bound to a method inspired by Fano's inequality where the continuous random variable is discretized.

Non-lambertian reflectance modeling and shape recovery of faces using tensor splines.

Modeling illumination effects and pose variations of a face is of fundamental importance in the field of facial image analysis. Most of the conventional techniques that simultaneously address both of these problems work with the Lambertian assumption and thus fall short of accurately capturing the complex intensity variation that the facial images exhibit or recovering their 3D shape in the presence of specularities and cast shadows. In this paper, we present a novel Tensor-Spline-based framework for facial image analysis. We show that, using this framework, the facial apparent BRDF field can be accurately estimated while seamlessly accounting for cast shadows and specularities. Further, using local neighborhood information, the same framework can be exploited to recover the 3D shape of the face (to handle pose variation). We quantitatively validate the accuracy of the Tensor Spline model using a more general model based on the mixture of single-lobed spherical functions. We demonstrate the effectiveness of our technique by presenting extensive experimental results for face relighting, 3D shape recovery, and face recognition using the Extended Yale B and CMU PIE benchmark data sets.

A method for compact image representation using sparse matrix and tensor projections onto exemplar orthonormal bases.

We present a new method for compact representation of large image datasets. Our method is based on treating small patches from a 2-D image as matrices as opposed to the conventional vectorial representation, and encoding these patches as sparse projections onto a set of exemplar orthonormal bases, which are learned a priori from a training set. The end result is a low-error, highly compact image/patch representation that has significant theoretical merits and compares favorably with existing techniques (including JPEG) on experiments involving the compression of ORL and Yale face databases, as well as a database of miscellaneous natural images. In the context of learning multiple orthonormal bases, we show the easy tunability of our method to efficiently represent patches of different complexities. Furthermore, we show that our method is extensible in a theoretically sound manner to higher-order matrices ("tensors"). We demonstrate applications of this theory to compression of well-known color image datasets such as the GaTech and CMU-PIE face databases and show performance competitive with JPEG. Lastly, we also analyze the effect of image noise on the performance of our compression schemes.

Probability density estimation using isocontours and isosurfaces: applications to information-theoretic image registration.

We present a new, geometric approach for determining the probability density of the intensity values in an image. We drop the notion of an image as a set of discrete pixels, and assume a piecewise-continuous representation. The probability density can then be regarded as being proportional to the area between two nearby isocontours of the image surface. Our paper extends this idea to joint densities of image pairs. We demonstrate the application of our method to affine registration between two or more images using information theoretic measures such as mutual information. We show cases where our method outperforms existing methods such as simple histograms, histograms with partial volume interpolation, Parzen windows, etc. under fine intensity quantization for affine image registration under significant image noise. Furthermore, we demonstrate results on simultaneous registration of multiple images, as well as for pairs of volume datasets, and show some theoretical properties of our density estimator. Our approach requires the selection of only an image interpolant. The method neither requires any kind of kernel functions (as in Parzen windows) which are unrelated to the structure of the image in itself, nor does it rely on any form of sampling for density estimation.

Dynamical constraints on using precise spike timing to compute in recurrent cortical networks.

Several recent models have proposed the use of precise timing of spikes for cortical computation. Such models rely on growing experimental evidence that neurons in the thalamus as well as many primary sensory cortical areas respond to stimuli with remarkable temporal precision. Models of computation based on spike timing, where the output of the network is a function not only of the input but also of an independently initializable internal state of the network, must, however, satisfy a critical constraint: the dynamics of the network should not be sensitive to initial conditions. We have previously developed an abstract dynamical system for networks of spiking neurons that has allowed us to identify the criterion for the stationary dynamics of a network to be sensitive to initial conditions. Guided by this criterion, we analyzed the dynamics of several recurrent cortical architectures, including one from the orientation selectivity literature. Based on the results, we conclude that under conditions of sustained, Poisson-like, weakly correlated, low to moderate levels of internal activity as found in the cortex, it is unlikely that recurrent cortical networks can robustly generate precise spike trajectories, that is, spatiotemporal patterns of spikes precise to the millisecond timescale.

Large Margin Pursuit for a Conic Section Classifier.

Learning a discriminant becomes substantially more difficult when the datasets are high-dimensional and the available samples are few. This is often the case in computer vision and medical diagnosis applications. A novel Conic Section classifier (CSC) was recently introduced in the literature to handle such datasets, wherein each class was represented by a conic section parameterized by its focus, directrix and eccentricity. The discriminant boundary was the locus of all points that are equi-eccentric relative to each class-representative conic section. Simpler boundaries were preferred for the sake of generalizability.In this paper, we improve the performance of the two-class classifier via a large margin pursuit. When formulated as a non-linear optimization problem, the margin computation is demonstrated to be hard, especially due to the high dimensionality of the data. Instead, we present a geometric algorithm to compute the distance of a point to the nonlinear discriminant boundary generated by the CSC in the input space. We then introduce a large margin pursuit in the learning phase so as to enhance the generalization capacity of the classifier. We validate the algorithm on real datasets and show favorable classification rates in comparison to many existing state-of-the-art binary classifiers as well as the CSC without margin pursuit.

BLISS 2.0: a web-based tool for predicting conserved regulatory modules in distantly-related orthologous sequences.

UNLABELLED: BLISS 2.0 is a web-based application for identifying conserved regulatory modules in distantly related orthologous sequences. Unlike existing approaches, it performs the cross-genome comparison at the binding site level. Experimental results on simulated and real world data indicate that BLISS 2.0 can identify conserved regulatory modules from sequences with little overall similarity at the DNA sequence level. AVAILABILITY: http://www.blisstool.org/

BLISS: binding site level identification of shared signal-modules in DNA regulatory sequences.

BACKGROUND: Regulatory modules are segments of the DNA that control particular aspects of gene expression. Their identification is therefore of great importance to the field of molecular genetics. Each module is composed of a distinct set of binding sites for specific transcription factors. Since experimental identification of regulatory modules is an arduous process, accurate computational techniques that supplement this process can be very beneficial. Functional modules are under selective pressure to be evolutionarily conserved. Most current approaches therefore attempt to detect conserved regulatory modules through similarity comparisons at the DNA sequence level. However, some regulatory modules, despite the conservation of their responsible binding sites, are embedded in sequences that have little overall similarity. RESULTS: In this study, we present a novel approach that detects conserved regulatory modules via comparisons at the binding site level. The technique compares the binding site profiles of orthologs and identifies those segments that have similar (not necessarily identical) profiles. The similarity measure is based on the inner product of transformed profiles, which takes into consideration the p values of binding sites as well as the potential shift of binding site positions. We tested this approach on simulated sequence pairs as well as real world examples. In both cases our technique was able to identify regulatory modules which could not to be identified using sequence-similarity based approaches such as rVista 2.0 and Blast. CONCLUSION: The results of our experiments demonstrate that, for sequences with little overall similarity at the DNA sequence level, it is still possible to identify conserved regulatory modules based solely on binding site profiles.

On the sensitive dependence on initial conditions of the dynamics of networks of spiking neurons.

We have previously formulated an abstract dynamical system for networks of spiking neurons and derived a formal result that identifies the criterion for its dynamics, without inputs, to be "sensitive to initial conditions". Since formal results are applicable only to the extent to which their assumptions are valid, we begin this article by demonstrating that the assumptions are indeed reasonable for a wide range of networks, particularly those that lack overarching structure. A notable aspect of the criterion is the finding that sensitivity does not necessarily arise from randomness of connectivity or of connection strengths, in networks. The criterion guides us to cases that decouple these aspects: we present two instructive examples of networks, one with random connectivity and connection strengths, yet whose dynamics is insensitive, and another with structured connectivity and connection strengths, yet whose dynamics is sensitive. We then argue based on the criterion and the gross electrophysiology of the cortex that the dynamics of cortical networks ought to be almost surely sensitive under conditions typically found there. We supplement this with two examples of networks modeling cortical columns with widely differing qualitative dynamics, yet with both exhibiting sensitive dependence. Next, we use the criterion to construct a network that undergoes bifurcation from sensitive dynamics to insensitive dynamics when the value of a control parameter is varied. Finally, we extend the formal result to networks driven by stationary input spike trains, deriving a superior criterion than previously reported.

A Conic Section Classifier and its Application to Image Datasets.

Many problems in computer vision involving recognition and/or classification can be posed in the general framework of supervised learning. There is however one aspect of image datasets, the high-dimensionality of the data points, that makes the direct application of off-the-shelf learning techniques problematic. In this paper, we present a novel concept class and a companion tractable algorithm for learning a suitable classifier from a given labeled dataset, that is particularly suited to high-dimensional sparse datasets. Each member class in the dataset is represented by a prototype conic section in the feature space, and new data points are classified based on a distance measure to each such representative conic section that is parameterized by its focus, directrix and eccentricity. Learning is achieved by altering the parameters of the conic section descriptor for each class, so as to better represent the data. We demonstrate the efficacy of the technique by comparing it to several well known classifiers on multiple public domain datasets.