Angular Embedding: A Robust Quadratic Criterion

SurReal: Complex-Valued Learning as Principled Transformations on a Scaling and Rotation Manifold

Complex-valued data are ubiquitous in signal and image processing applications, and complex-valued representations in deep learning have appealing theoretical properties. While these aspects have long been recognized, complex-valued deep learning continues to lag far behind its real-valued counterpart. We propose a principled geometric approach to complex-valued deep learning. Complex-valued data could often be subject to arbitrary complex-valued scaling; as a result, real and imaginary components could covary. Instead of treating complex values as two independent channels of real values, we recognize their underlying geometry: we model the space of complex numbers as a product manifold of nonzero scaling and planar rotations. Arbitrary complex-valued scaling naturally becomes a group of transitive actions on this manifold. We propose to extend the property instead of the form of real-valued functions to the complex domain. We define convolution as the weighted Fréchet mean on the manifold that is equivariant to the group of scaling/rotation actions and define distance transform on the manifold that is invariant to the action group. The manifold perspective also allows us to define nonlinear activation functions, such as tangent ReLU and G -transport, as well as residual connections on the manifold-valued data. We dub our model SurReal, as our experiments on MSTAR and RadioML deliver high performance with only a fractional size of real- and complex-valued baseline models.

A computational strategy to understand structure-activity relationship of 1,3-disubstituted imidazole [1,5-α] pyrazine derivatives described as ATP competitive inhibitors of the IGF-1 receptor related to Ewing sarcoma

We followed a comprehensive computational strategy to understand and eventually predict the structure-activity relationship of thirty-three 1,3-disubstituted imidazole [1,5-α] pyrazine derivatives described as ATP competitive inhibitors of the IGF-1 receptor related to Ewing sarcoma. The quantitative structure-activity relationship model showed that the inhibitory potency is correlated with the molar volume, a steric descriptor and the net charge calculated value on atom C1 (q1) and N4 (q4) of the pharmacophore, all of them appearing to give a positive contribution to the inhibitory activity. According to experimental and calculated values, the most potent compound would be 3-[4-(azetidin-2-ylmethyl) cyclohexyl]-1-[3-(benzyloxy) phenyl] imidazo [1,5-α]pyrazin-8-amine (compound 23). Docking was used to guess important residues involved in the ATP-competitive inhibitory activity. It was validated by 200 ns of molecular dynamics (MD) simulation using improved linear interaction energy (LIE) method. MD of previously preferred structures by docking shows that the most potent ligand could establish hydrogen bonds with the ATP-binding site of the receptor, and the Ser979 and Ser1059 residues contribute favourably to the binding stability of compound 23. MD simulation also gave arguments about the chemical structure of the compound 23 being able to fit in the ATP-binding pocket, expecting to remain stable into it during the entire simulation and allowing us to hint the significant contribution expected to be given by electrostatic and hydrophobic interactions to the ligand-receptor complex stability. This computational combined strategy here described could represent a useful and effective prime approach to guide the identification of tyrosine kinase inhibitors as new lead compounds.

From Social to Individuals: A Parsimonious Path of Multi-Level Models for Crowdsourced Preference Aggregation

In crowdsourced preference aggregation, it is often assumed that all the annotators are subject to a common preference or social utility function which generates their comparison behaviors in experiments. However, in reality, annotators are subject to variations due to multi-criteria, abnormal, or a mixture of such behaviors. In this paper, we propose a parsimonious mixed-effects model, which takes into account both the fixed effect that the majority of annotators follows a common linear utility model, and the random effect that some annotators might deviate from the common significantly and exhibit strongly personalized preferences. The key algorithm in this paper establishes a dynamic path from the social utility to individual variations, with different levels of sparsity on personalization. The algorithm is based on the Linearized Bregman Iterations, which leads to easy parallel implementations to meet the need of large-scale data analysis. In this unified framework, three kinds of random utility models are presented, including the basic linear model with L₂ loss, Bradley-Terry model, and Thurstone-Mosteller model. The validity of these multi-level models are supported by experiments with both simulated and real-world datasets, which shows that the parsimonious multi-level models exhibit improvements in both interpretability and predictive precision compared with traditional HodgeRank.

Saliency Detection via Absorbing Markov Chain With Learnt Transition Probability

In this paper, we propose a bottom-up saliency model based on absorbing Markov chain (AMC). First, a sparsely connected graph is constructed to capture the local context information of each node. All image boundary nodes and other nodes are, respectively, treated as the absorbing nodes and transient nodes in the absorbing Markov chain. Then, the expected number of times from each transient node to all other transient nodes can be used to represent the saliency value of this node. The absorbed time depends on the weights on the path and their spatial coordinates, which are completely encoded in the transition probability matrix. Considering the importance of this matrix, we adopt different hierarchies of deep features extracted from fully convolutional networks and learn a transition probability matrix, which is called learnt transition probability matrix. Although the performance is significantly promoted, salient objects are not uniformly highlighted very well. To solve this problem, an angular embedding technique is investigated to refine the saliency results. Based on pairwise local orderings, which are produced by the saliency maps of AMC and boundary maps, we rearrange the global orderings (saliency value) of all nodes. Extensive experiments demonstrate that the proposed algorithm outperforms the state-of-the-art methods on six publicly available benchmark data sets.

Correction of geometrically distorted underwater images using shift map analysis

In underwater imaging, water waves cause severe geometric distortions and blurring of the acquired short-exposure images. Corrections for these distortions have been tackled reasonably well by previous efforts but still need improvement in the estimation of pixel shift maps to increase restoration accuracy. This paper presents a new algorithm that efficiently estimates the shift maps from geometrically distorted video sequences and uses those maps to restore the sequences. A nonrigid image registration method is employed to estimate the shift maps of the distorted frames against a reference frame. The sharpest frame of the sequence, determined using a sharpness metric, is chosen as the reference frame. A k-means clustering technique is employed to discard too-blurry frames that could result in inaccuracy in the shift maps' estimation. The estimated pixel shift maps are processed to generate the accurate shift map that is used to dewarp the input frames into their nondistorted forms. The proposed method is applied on several synthetic and real-world video sequences, and the obtained results exhibit significant improvements over the state-of-the-art methods.

Restoring atmospheric-turbulence-degraded images

Image data experiences geometric distortions and spatial-temporal varying blur due to the strong effects of random spatial and temporal variations in the optical refractive index of the communication path. Simultaneously removing these effects from an image is a challenging task. An efficient approach is proposed in this paper to address this problem. The approach consists of four steps. First, a frame selection strategy is employed by proposing an unsupervised k-means clustering technique. Second, a B-spline-based nonrigid image registration is carried out to suppress geometric distortions. Third, a spatiotemporal kernel regression is proposed by introducing the local sharp patch concept to fuse the registered frame sequences into an image. Finally, a blind deconvolution technique is employed to deblur the fused image. Experiments are carried out with synthetic and real-world turbulence-degraded data by implementing the proposed method and two recently reported methods. The proposed method demonstrates significant improvement over the two reported methods in terms of alleviating blur and distortions, as well as improving visual quality.

Robust Subjective Visual Property Prediction from Crowdsourced Pairwise Labels

The problem of estimating subjective visual properties from image and video has attracted increasing interest. A subjective visual property is useful either on its own (e.g. image and video interestingness) or as an intermediate representation for visual recognition (e.g. a relative attribute). Due to its ambiguous nature, annotating the value of a subjective visual property for learning a prediction model is challenging. To make the annotation more reliable, recent studies employ crowdsourcing tools to collect pairwise comparison labels. However, using crowdsourced data also introduces outliers. Existing methods rely on majority voting to prune the annotation outliers/errors. They thus require a large amount of pairwise labels to be collected. More importantly as a local outlier detection method, majority voting is ineffective in identifying outliers that can cause global ranking inconsistencies. In this paper, we propose a more principled way to identify annotation outliers by formulating the subjective visual property prediction task as a unified robust learning to rank problem, tackling both the outlier detection and learning to rank jointly. This differs from existing methods in that (1) the proposed method integrates local pairwise comparison labels together to minimise a cost that corresponds to global inconsistency of ranking order, and (2) the outlier detection and learning to rank problems are solved jointly. This not only leads to better detection of annotation outliers but also enables learning with extremely sparse annotations.

Differentiation of benign pigmented skin lesions with the aid of computer image analysis: a novel approach

BACKGROUND

The differential diagnosis of common pigmented skin lesions is important in cosmetic dermatology. The computer aided image analysis would be a potent ancillary diagnostic tool when patients are hesitant to undergo a skin biopsy.

OBJECTIVE

We investigated the numerical parameters discriminating each pigmented skin lesion from another with statistical significance.

METHODS

For each of the five magnified digital images containing clinically diagnosed nevus, lentigo and seborrheic keratosis, a total of 23 parameters describing the morphological, color, texture and topological features were calculated with the aid of a self-developed image analysis software. A novel concept of concentricity was proposed, which represents how closely the color segmentation resembles a concentric circle.

RESULTS

Morphologically, seborrheic keratosis was bigger and spikier than nevus and lentigo. The color histogram revealed that nevus was the darkest and had the widest variation in tone. In the aspect of texture, the surface of the nevus showed the highest contrast and correlation. Finally, the color segmented pattern of the nevus and lentigo was far more concentric than that of seborrheic keratosis.

CONCLUSION

We found that the subtle distinctions between nevus, lentigo and seborrheic keratosis, which are likely to be unrecognized by ocular inspection, are well emphasized and detected with the aid of software.