Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Computational single-cell RNA-seq (scRNA-seq) methods have been successfully applied to experiments representing a single condition, technology, or species to discover and define cellular phenotypes. However, identifying subpopulations of cells that are present across multiple data sets remains challenging. Here, we introduce an analytical strategy for integrating scRNA-seq data sets based on common sources of variation, enabling the identification of shared populations across data sets and downstream comparative analysis. We apply this approach, implemented in our R toolkit Seurat (http://satijalab.org/seurat/), to align scRNA-seq data sets of peripheral blood mononuclear cells under resting and stimulated conditions, hematopoietic progenitors sequenced using two profiling technologies, and pancreatic cell 'atlases' generated from human and mouse islets. In each case, we learn distinct or transitional cell states jointly across data sets, while boosting statistical power through integrated analysis. Our approach facilitates general comparisons of scRNA-seq data sets, potentially deepening our understanding of how distinct cell states respond to perturbation, disease, and evolution.

/pdf/integrating-single-cell-transcriptomic-data-across-different-3dtsi5w7ob.pdf

Integrating single-cell transcriptomic data across different conditions, technologies, and species.

/pdf/convex-analysisnoer-san-nojin-zhan-nituite-5dl2rbmoyr.pdf

Convex Analysisの二,三の進展について

This paper presents a novel and efficient facial image representation based on local binary pattern (LBP) texture features. The face image is divided into several regions from which the LBP feature distributions are extracted and concatenated into an enhanced feature vector to be used as a face descriptor. The performance of the proposed method is assessed in the face recognition problem under different challenges. Other applications and several extensions are also discussed

Face Description with Local Binary Patterns: Application to Face Recognition

“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

Video face recognition is a challenging recognition problem due to low-quality and redundant video data. In order to efficiently address the problem, this paper proposes an adversarial embedding and variational aggregation (AEVA) approach that takes discriminative information from adversarial learning to implicitly constrain the distributions of video features. AEVA contains two parts: adversarial embedding learning and variational aggregation learning. The former contributes to discriminative feature embedding space by a self-adversarial process. It reduces intra-class distance of the same subject in an adversarial way. The latter aims to improve aggregated video representation by forcing latent feature distribution to be as close to real feature distribution as possible. An attentional weighting network and a variational inference structure are used to aggregate the features from one video and generate the latent features. Both parts have neither complex sampling strategies nor the hyperparameter settings from deep metric learning. Our approach achieves state-of-the-art performance for video face recognition on four widely used benchmarks, including YouTubeFace, IJB-A, YouTube Celebrities and Celebrity-1000.

Adversarial Embedding and Variational Aggregation for Video Face Recognition

<h3>Purpose/Objective(s)</h3> Sarcopenia is a well-known prognostic factor in head and neck cancer (HNC) patients. Sarcopenia status in HNC patients can be determined using the skeletal muscle (SM) index (SMI) as calculated from the cross-sectional area (CSA) of C3 cervical vertebrae musculature. However, current approaches to generate the C3 SM segmentations rely on human judgment which is time consuming and prone to inter-observer variability. We hypothesize that deep learning (DL) can be used to develop a fully automated approach to segment SM at the C3 vertebral level for SMI calculations which can be used for subsequent body composition and sarcopenia assessment. <h3>Materials/Methods</h3> 390 HNC patients from a public database with corresponding contrast-enhanced computed tomography (CT) scans were utilized. Patients were split into training (n=300) and testing (n=90) sets. Ground truth (GT) single-slice SM segmentations at the C3 vertebral level were manually generated. A multi-stage DL pipeline was developed using the MONAI package, where a 3D ResUnet model auto-segmented the C3 section, the middle slice of the segmented C3 section was auto-selected, and a 2D ResUnet model auto-segmented the auto-selected slice. Both the 3D and 2D approaches trained five sub-models (5-fold cross-validation) and combined sub-model predictions on the test set using majority vote ensembling. Model performance was determined using the Dice similarity coefficient (DSC). Additionally, we quantified the absolute difference between the slice locations of the C3 mid-section predicted by the 3D model and the single-slice CT GT image. Furthermore, we compared the GT C3 vertebral level SM CSA and the corresponding SMI to the 2D model predictions. Finally, using these SMI values coupled to previously established sarcopenia stratification cutoffs, we performed a Kaplan Meier analysis on the entire dataset to determine associations to overall survival. <h3>Results</h3> Mean test set DSC of the 3D and 2D models were 0.96 and 0.95, respectively. The mid-slices of the 3D model segmentations were within 2 mm distance of the GT slices for 78% of the test patients. The CSA derived from the 2D model predictions using the auto-selected slice from the 3D model and corresponding SMI values were highly correlated to the CSA (r=0.98) and SMI (r=0.96) derived from the GT segmentations. Kaplan Meier analysis revealed that the SMI values derived from auto-segmented SM could stratify patients for overall survival in males (log-rank p < 0.05) but not females (log-rank p > 0.05), with results that were consistent with SMI values derived from the GT segmentations. <h3>Conclusion</h3> Using open-source toolkits and public datasets, we developed a high-performance multi-stage DL auto-segmentation for SM at the C3 vertebrae. This work is a vital first step towards fully automated workflows for sarcopenia-related clinical decision making. 

Cervical Vertebrae Skeletal Muscle Auto Segmentation for Sarcopenia Analysis Using Pre-Therapy CT in Head and Neck Cancer Patients

Identity transfer often faces the challenge of generalizing to new situations where large pose and expression or background gaps exist between source and target face images. To improve generalization in such situations, biases take a key role~\cite{mitchell_1980_bias}. This paper proposes an Errors-in-Variables Adapter (EVA) model to induce learning of proper generalizations by explicitly employing biases to identity estimation based on prior knowledge about the target situation. To better match the source face with the target situation in terms of pose, expression, and background factors, we model the bias as a causal effect of the target situation on source identity and estimate this effect through a controlled intervention trial. To achieve smoother transfer for the target face across the identity gap, we eliminate the target face specificity through multiple kernel regressions. The kernels are used to constrain the regressions to operate only on identity information in the internal representations of the target image, while leaving other perceptual information invariant. Combining these post-regression representations with the biased estimation for identity, EVA shows impressive performance even in the presence of large gaps, providing empirical evidence supporting the utility of the inductive biases in identity estimation.

Inductive Biased Estimation: Learning Generalizations for Identity Transfer

Real-world image super-resolution (SR) is a challenging image translation problem. Low-resolution (LR) images are often generated by various unknown transformations rather than by applying simple bilinear down-sampling on high-resolution (HR) images. To address this issue, this paper proposes a novel pipeline which exploits style and attention mechanism in real-world SR. Our pipeline consists of a style Variational Autoencoder (styleVAE) and a SR network incorporated with attention mechanism. To get real-world-like low-quality images paired with the HR images, we design the styleVAE to transfer the complex nuisance factors in real-world LR images to the generated LR images. We also use mutual information estimation (MI) to get better style information. For our SR network, we firstly propose a global attention residual block to learn long-range dependencies in images. Then another local attention residual block is proposed to enforce the attention of SR network moving to local areas of images in which texture detail will be filled. It is worth noticing that styleVAE can be presented in a plug-and-play manner and thus can help to improve the generalization and robustness of our SR method as well as other SR methods. Extensive experiments demonstrate that our method surpasses the state-of-the-art work, both quantitatively and qualitatively.

Exploiting Style and Attention in Real-World Super-Resolution

With the growing interest in pretrained vision-language models like CLIP, recent research has focused on adapting these models to downstream tasks. Despite achieving promising results, most existing methods require labeled data for all classes, which may not hold in real-world applications due to the long tail and Zipf's law. For example, some classes may lack labeled data entirely, such as emerging concepts. To address this problem, we propose a plug-and-play generative approach called \textbf{S}ynt\textbf{H}es\textbf{I}zed \textbf{P}rompts~(\textbf{SHIP}) to improve existing fine-tuning methods. Specifically, we follow variational autoencoders to introduce a generator that reconstructs the visual features by inputting the synthesized prompts and the corresponding class names to the textual encoder of CLIP. In this manner, we easily obtain the synthesized features for the remaining label-only classes. Thereafter, we fine-tune CLIP with off-the-shelf methods by combining labeled and synthesized features. Extensive experiments on base-to-new generalization, cross-dataset transfer learning, and generalized zero-shot learning demonstrate the superiority of our approach. The code is available at \url{https://github.com/mrflogs/SHIP}.

Ran He

Papers

Adversarial Embedding and Variational Aggregation for Video Face Recognition

Cervical Vertebrae Skeletal Muscle Auto Segmentation for Sarcopenia Analysis Using Pre-Therapy CT in Head and Neck Cancer Patients

Inductive Biased Estimation: Learning Generalizations for Identity Transfer

Exploiting Style and Attention in Real-World Super-Resolution

Improving Zero-Shot Generalization for CLIP with Synthesized Prompts