Search for: All records

Interpreting deep neural networks through examining neurons offers distinct advantages when it comes to exploring the inner workings of Deep Neural Networks. Previous research has indicated that specific neurons within deep vision networks possess semantic meaning and play pivotal roles in model performance. Nonetheless, the current methods for generating neuron semantics heavily rely on human intervention, which hampers their scalability and applicability. To address this limitation, this paper proposes a novel post-hoc framework for generating semantic explanations of neurons with large foundation models, without requiring human intervention or prior knowledge. Experiments are conducted with both qualitative and quantitative analysis to verify the effectiveness of our proposed approach.

 

Hypergraph neural networks (HyperGNNs) are a family of deep neural networks designed to perform inference on hypergraphs. HyperGNNs follow either a spectral or a spatial approach, in which a convolution or message-passing operation is conducted based on a hypergraph algebraic descriptor. While many HyperGNNs have been proposed and achieved state-of-the-art performance on broad applications, there have been limited attempts at exploring high-dimensional hypergraph descriptors (tensors) and joint node interactions carried by hyperedges. In this article, we depart from hypergraph matrix representations and present a new tensor-HyperGNN (T-HyperGNN) framework with cross-node interactions (CNIs). The T-HyperGNN framework consists of T-spectral convolution, T-spatial convolution, and T-message-passing HyperGNNs (T-MPHN). The T-spectral convolution HyperGNN is defined under the t-product algebra that closely connects to the spectral space. To improve computational efficiency for large hypergraphs, we localize the T-spectral convolution approach to formulate the T-spatial convolution and further devise a novel tensor-message-passing algorithm for practical implementation by studying a compressed adjacency tensor representation. Compared to the state-of-the-art approaches, our T-HyperGNNs preserve intrinsic high-order network structures without any hypergraph reduction and model the joint effects of nodes through a CNI layer. These advantages of our T-HyperGNNs are demonstrated in a wide range of real-world hypergraph datasets. The implementation code is available at https://github.com/wangfuli/T-HyperGNNs.git. 

Functional magnetic resonance imaging faces inherent challenges when applied to deep-brain areas in rodents, e.g. entorhinal cortex, due to the signal loss near the ear cavities induced by susceptibility artifacts and reduced sensitivity induced by the long distance from the surface array coil. Given the pivotal roles of deep brain regions in various diseases, optimized imaging techniques are needed. To mitigate susceptibility-induced signal losses, we introduced baby cream into the middle ear. To enhance the detection sensitivity of deep brain regions, we implemented inductively coupled ear-bars, resulting in approximately a 2-fold increase in sensitivity in entorhinal cortex. Notably, the inductively coupled ear-bar can be seamlessly integrated as an add-on device, without necessitating modifications to the scanner interface. To underscore the versatility of inductively coupled ear-bars, we conducted echo-planner imaging-based task functional magnetic resonance imaging in rats modeling Alzheimer’s disease. As a proof of concept, we also demonstrated resting-state-functional magnetic resonance imaging connectivity maps originating from the left entorhinal cortex—a central hub for memory and navigation networks-to amygdala hippocampal area, Insular Cortex, Prelimbic Systems, Cingulate Cortex, Secondary Visual Cortex, and Motor Cortex. This work demonstrates an optimized procedure for acquiring large-scale networks emanating from a previously challenging seed region by conventional magnetic resonance imaging detectors, thereby facilitating improved observation of functional magnetic resonance imaging outcomes.

 

Given the availability of abundant data, deep learning models have been advanced and become ubiquitous in the past decade. In practice, due to many different reasons (e.g., privacy, usability, and fidelity), individuals also want the trained deep models to forget some specific data. Motivated by this, machine unlearning (also known as selective data forgetting) has been intensively studied, which aims at removing the influence that any particular training sample had on the trained model during the unlearning process. However, people usually employ machine unlearning methods as trusted basic tools and rarely have any doubt about their reliability. In fact, the increasingly critical role of machine unlearning makes deep learning models susceptible to the risk of being maliciously attacked. To well understand the performance of deep learning models in malicious environments, we believe that it is critical to study the robustness of deep learning models to malicious unlearning attacks, which happen during the unlearning process. To bridge this gap, in this paper, we first demonstrate that malicious unlearning attacks pose immense threats to the security of deep learning systems. Specifically, we present a broad class of malicious unlearning attacks wherein maliciously crafted unlearning requests trigger deep learning models to misbehave on target samples in a highly controllable and predictable manner. In addition, to improve the robustness of deep learning models, we also present a general defense mechanism, which aims to identify and unlearn effective malicious unlearning requests based on their gradient influence on the unlearned models. Further, theoretical analyses are conducted to analyze the proposed methods. Extensive experiments on real-world datasets validate the vulnerabilities of deep learning models to malicious unlearning attacks and the effectiveness of the introduced defense mechanism. 

Polyploidy is a major evolutionary force that has shaped plant diversity. However, the various pathways toward polyploid formation and interploidy gene flow remain poorly understood. Here, we demonstrated that the immediate progeny of allotriploid AACBrassica(obtained by crossing allotetraploidBrassica napusand diploidBrassica rapa) was predominantly aneuploids with ploidal levels ranging from near-triploidy to near-hexaploidy, and their chromosome numbers deviated from the theoretical distribution toward increasing chromosome numbers, suggesting that they underwent selection. Karyotype and phenotype analyses showed that aneuploid individuals containing fewer imbalanced chromosomes had higher viability and fertility. Within three generations of self-fertilization, allotriploids mainly developed into near or complete allotetraploids similar toB. napusvia gradually increasing chromosome numbers and fertility, suggesting that allotriploids could act as a bridge in polyploid formation, with aneuploids as intermediates. Self-fertilized interploidy hybrids ultimately generated new allopolyploids carrying different chromosome combinations, which may create a reproductive barrier preventing allotetraploidy back to diploidy and promote gene flow from diploids to allotetraploids. These results suggest that the maintenance of a proper genome balance and dosage drove the recurrent conversion of allotriploids to allotetraploids, which may contribute to the formation and evolution of polyploids.