arXiv subjects

Zihao Wu

Publications and source records attributed to Zihao Wu.

151 records · Page 9Linked to original sources

arXiv2022

Coupling Visual Semantics of Artificial Neural Networks and Human Brain Function via Synchronized Activations

Artificial neural networks (ANNs), originally inspired by biological neural networks (BNNs), have achieved remarkable successes in many tasks such as visual representation learning. However, whether there exists semantic correlations/connections between the visual representations in ANNs and those in BNNs remains largely unexplored due to both the lack of an effective tool to link and couple two different domains, and the lack of a general and effective framework of representing the visual semantics in BNNs such as human functional brain networks (FBNs). To answer this question, we propose a novel computational framework, Synchronized Activations (Sync-ACT), to couple the visual representation spaces and semantics between ANNs and BNNs in human brain based on naturalistic functional magnetic resonance imaging (nfMRI) data. With this approach, we are able to semantically annotate the neurons in ANNs with biologically meaningful description derived from human brain imaging for the first time. We evaluated the Sync-ACT framework on two publicly available movie-watching nfMRI datasets. The experiments demonstrate a) the significant correlation and similarity of the semantics between the visual representations in FBNs and those in a variety of convolutional neural networks (CNNs) models; b) the close relationship between CNN's visual representation similarity to BNNs and its performance in image classification tasks. Overall, our study introduces a general and effective paradigm to couple the ANNs and BNNs and provides novel insights for future studies such as brain-inspired artificial intelligence.

cs.CV↗

arXiv2022

On Schwinger pair production between D3 branes

We study the open string pair production between two D3 branes, which will give rise to similar effect as Schwinger pair production for observers on one of the D3 branes. The D3 branes are placed parallel at a distance, and they are carrying world-volume electromagnetic fluxes that takes general form. We derive the pair production rate by computing the interaction amplitude between the D3 branes. We discussed how to maximize the pair production rate in this general case. We also mentioned that the general result can be used to describe other system such as D3-D1, where the pair production is ultra large compared to original Schwinger pair production, making it hopeful to observe pair production in experiments.

hep-th↗

arXiv2022

Eye-gaze-guided Vision Transformer for Rectifying Shortcut Learning

Learning harmful shortcuts such as spurious correlations and biases prevents deep neural networks from learning the meaningful and useful representations, thus jeopardizing the generalizability and interpretability of the learned representation. The situation becomes even more serious in medical imaging, where the clinical data (e.g., MR images with pathology) are limited and scarce while the reliability, generalizability and transparency of the learned model are highly required. To address this problem, we propose to infuse human experts' intelligence and domain knowledge into the training of deep neural networks. The core idea is that we infuse the visual attention information from expert radiologists to proactively guide the deep model to focus on regions with potential pathology and avoid being trapped in learning harmful shortcuts. To do so, we propose a novel eye-gaze-guided vision transformer (EG-ViT) for diagnosis with limited medical image data. We mask the input image patches that are out of the radiologists' interest and add an additional residual connection in the last encoder layer of EG-ViT to maintain the correlations of all patches. The experiments on two public datasets of INbreast and SIIM-ACR demonstrate our EG-ViT model can effectively learn/transfer experts' domain knowledge and achieve much better performance than baselines. Meanwhile, it successfully rectifies the harmful shortcut learning and significantly improves the EG-ViT model's interpretability. In general, EG-ViT takes the advantages of both human expert's prior knowledge and the power of deep neural networks. This work opens new avenues for advancing current artificial intelligence paradigms by infusing human intelligence.

cs.CV↗

arXiv2022

Mask-guided Vision Transformer (MG-ViT) for Few-Shot Learning

Learning with little data is challenging but often inevitable in various application scenarios where the labeled data is limited and costly. Recently, few-shot learning (FSL) gained increasing attention because of its generalizability of prior knowledge to new tasks that contain only a few samples. However, for data-intensive models such as vision transformer (ViT), current fine-tuning based FSL approaches are inefficient in knowledge generalization and thus degenerate the downstream task performances. In this paper, we propose a novel mask-guided vision transformer (MG-ViT) to achieve an effective and efficient FSL on ViT model. The key idea is to apply a mask on image patches to screen out the task-irrelevant ones and to guide the ViT to focus on task-relevant and discriminative patches during FSL. Particularly, MG-ViT only introduces an additional mask operation and a residual connection, enabling the inheritance of parameters from pre-trained ViT without any other cost. To optimally select representative few-shot samples, we also include an active learning based sample selection method to further improve the generalizability of MG-ViT based FSL. We evaluate the proposed MG-ViT on both Agri-ImageNet classification task and ACFR apple detection task with gradient-weighted class activation mapping (Grad-CAM) as the mask. The experimental results show that the MG-ViT model significantly improves the performance when compared with general fine-tuning based ViT models, providing novel insights and a concrete approach towards generalizing data-intensive and large-scale deep learning models for FSL.

cs.CV↗

arXiv2020

IBP reduction coefficients made simple

We present an efficient method to shorten the analytic integration-by-parts (IBP) reduction coefficients of multi-loop Feynman integrals. For our approach, we develop an improved version of Leinartas' multivariate partial fraction algorithm, and provide a modern implementation based on the computer algebra system Singular. Furthermore, We observe that for an integral basis with uniform transcendental (UT) weights, the denominators of IBP reduction coefficients with respect to the UT basis are either symbol letters or polynomials purely in the spacetime dimension $D$. With a UT basis, the partial fraction algorithm is more efficient both with respect to its performance and the size reduction. We show that in complicated examples with existence of a UT basis, the IBP reduction coefficients size can be reduced by a factor of as large as $\sim 100$. We observe that our algorithm also works well for settings without a UT basis.

hep-ph↗

arXiv2020

On D-brane interaction & its related properties

We compute the closed-string cylinder amplitude between one Dp brane and the other Dp$\prime$ brane, placed parallel at a separation, with each carrying a general constant worldvolume flux and with $p - p' = 0, 2, 4, 6$ and $p \le 6$. For the $p = p'$, we show that the main part of the amplitude for $p = p' < 5$ is a special case of that for $p = p' = 5$ or $6$ case. For all other $p - p' = 2, 4, 6$ cases, we show that the amplitude is just a special case of the corresponding one for $p = p'$ case. Combining both, we obtain the general formula for the amplitude, which is valid for each of the cases considered and for arbitrary constant worldvolume fluxes. The corresponding general open string one-loop annulus amplitude is also obtained by a Jacobi transformation of the general cylinder one. We give also the general open string pair production rate. We study the properties of the amplitude such as the nature of the interaction, the open string tachyonic instability, and the possible open string pair production and its potential enhancement. In particular, in the presence of pure magnetic fluxes or magnetic-like fluxes, we find that the nature of interaction is correlated with the existence of potential open string tachyonic instability. When the interaction is attractive, there always exists an open string tachyonic instability when the brane separation reaches the minimum determined by the so-called tachyonic shift. When the interaction is repulsive, there is no such instability for any brane separation. We also find that the enhancement of open string pair production, in the presence of pure electric fluxes, can occur only for the $p - p' = 2$ case.

hep-th↗

arXiv2019

Extracting 2D weak labels from volume labels using multiple instance learning in CT hemorrhage detection

Multiple instance learning (MIL) is a supervised learning methodology that aims to allow models to learn instance class labels from bag class labels, where a bag is defined to contain multiple instances. MIL is gaining traction for learning from weak labels but has not been widely applied to 3D medical imaging. MIL is well-suited to clinical CT acquisitions since (1) the highly anisotropic voxels hinder application of traditional 3D networks and (2) patch-based networks have limited ability to learn whole volume labels. In this work, we apply MIL with a deep convolutional neural network to identify whether clinical CT head image volumes possess one or more large hemorrhages (> 20cm$^3$), resulting in a learned 2D model without the need for 2D slice annotations. Individual image volumes are considered separate bags, and the slices in each volume are instances. Such a framework sets the stage for incorporating information obtained in clinical reports to help train a 2D segmentation approach. Within this context, we evaluate the data requirements to enable generalization of MIL by varying the amount of training data. Our results show that a training size of at least 400 patient image volumes was needed to achieve accurate per-slice hemorrhage detection. Over a five-fold cross-validation, the leading model, which made use of the maximum number of training volumes, had an average true positive rate of 98.10%, an average true negative rate of 99.36%, and an average precision of 0.9698. The models have been made available along with source code to enabled continued exploration and adaption of MIL in CT neuroimaging.

cs.CV↗