Published:
A graph is a powerful way of representing relationships among entities as nodes connected by edges. Sometimes nodes and edges can have spatial features, such as 3D coordinates of nodes and directions along edges. How do we reason over the topology of graphs while considering those geometric features? Read more
Published:
GraphSAGE is a general inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data Read more
Published:
This loss function is a modification of the average Hausdorff distance between two unordered sets of points. The proposed method has no notion of bounding boxes, region proposals, or sliding windows. Read more
Published:
The origin ensemble (OE) algorithm is a novel statistical method for MMSE reconstruction of emission tomography data. This method allows one to perform reconstruction entirely in the image domain, i.e. without the use of forward and backprojection operations. Read more
Published:
In probability theory, the Gillespie algorithm generates a statistically correct trajectory (possible solution) of a stochastic equation system for which the reaction rates are known. Mathematically, it is a variant of a dynamic Monte Carlo method and similar to the kinetic Monte Carlo methods. Read more
Published:
The origin ensemble (OE) algorithm is a novel statistical method for MMSE reconstruction of emission tomography data. This method allows one to perform reconstruction entirely in the image domain, i.e. without the use of forward and backprojection operations. Read more
Published:
The Radon transform is the transform of our n-dimensional volume to a complete set of (n-1)-dimensional line integrals. Whereas the inverse Radon transform is the transform from our complete (n-1)-dimensional line integrals back to the original image. Read more
Published:
This paper explores the causalities among image features, contexts, and categories to eliminate the biased sobject-context entanglement in the class activation maps thus improving the accuracy of object localization. Read more
Published:
Long time no see. After a tiring semester, I finally have the chance to update posts. Though there is much research progress to be talked about, it demands some more time to collect my thoughts about it. So this post will be a rather simple introduction to a famous equation in both physics and computer graphics, the radiative transfer equation (RTE). Read more
Published:
Conventional trials of physics-informed deep learning (say, PINNs) integrates the dynamics as constraints to guide the approximation of a certain function within. An obvious problem with this idea is that such models cannot be directly applied to different initial and boundary conditions and observations. DeepONet, representing a new trial in physics-informed learning, employed some tricks to get around with this. Read more
Published:
Using in-distribution (ID) accuracy to analyze out-of-distribution (OOD) performance of models can encounter severe disintegrity, especially for models trained with diverse supervision and distributions, e.g., vision models (VM) and vision-language models (VLM). This paper, accepted at ICML 20241, proposed a novel, yet simple approach to measure OOD performance without introduction of any OOD data.
Published:
Integro-Differential Equations, IDEs for short, are equations involving both partial differential and integral operators. They are commonly used in a variety of disciplines of science and engineering topics, while nonlinear IDEs are particularly difficult to solve analytically. Therefore, the idea to incorporate neural networks arises naturally. This post is dedicated to recording my understanding of IDEs and NNs. Read more
Published:
Generative models, powered by AI, is a big trend now for stuff synthesis (texts, images, and even more specialize usages). This post probes into why and how they are working from a mathematical perspective. Read more
Published:
Thin, fragile local structures, and variable global morphologies, etc., of topological tubular structures like vessels and roads, lead to the difficulty of its accurate segmentation. This paper of DSCNet simultaneously enhance its perception for such tubular structures in 3 stages: feature extraction, feature fusion, and loss constraint. Read more
Published:
In the scenario of low-dose CT image reconstruction, supervised deep learning is widely adopted, which generally demands a dataset containing pairs of normal-dose and correspoding low-dose images and is therefore challenging in clinical situations. This paper proposed an unsupervised deep learning method to tackle this problem. Read more
Published:
总体而言，我们从两个方面改进了之前的Deeplab V3+ & DANN分割模型 Read more
Published:
Mainstream AIGC models, like ChatGPT and Stable Diffusion, have been drawing much attention over the recent years. Despite the massive amount of high-quality dataset, advanced network structure and loads of GPUs, one innegligible factor that contributes to their success is the use of RLHF, or reinforcement learning from human feedback. This blog is mostly based on a paper in May 2023 by the Levine Lab at UCB, and probes into how RLHF can be elegantly integrated into diffusion framework. Read more
Published:
总体而言，我们从两个方面改进了之前的Deeplab V3+ & DANN分割模型 Read more
Published:
Using in-distribution (ID) accuracy to analyze out-of-distribution (OOD) performance of models can encounter severe disintegrity, especially for models trained with diverse supervision and distributions, e.g., vision models (VM) and vision-language models (VLM). This paper, accepted at ICML 20241, proposed a novel, yet simple approach to measure OOD performance without introduction of any OOD data.
Published:
Mainstream AIGC models, like ChatGPT and Stable Diffusion, have been drawing much attention over the recent years. Despite the massive amount of high-quality dataset, advanced network structure and loads of GPUs, one innegligible factor that contributes to their success is the use of RLHF, or reinforcement learning from human feedback. This blog is mostly based on a paper in May 2023 by the Levine Lab at UCB, and probes into how RLHF can be elegantly integrated into diffusion framework. Read more
Published:
Generative models, powered by AI, is a big trend now for stuff synthesis (texts, images, and even more specialize usages). This post probes into why and how they are working from a mathematical perspective. Read more
Published:
A graph is a powerful way of representing relationships among entities as nodes connected by edges. Sometimes nodes and edges can have spatial features, such as 3D coordinates of nodes and directions along edges. How do we reason over the topology of graphs while considering those geometric features? Read more
Published:
A graph is a powerful way of representing relationships among entities as nodes connected by edges. Sometimes nodes and edges can have spatial features, such as 3D coordinates of nodes and directions along edges. How do we reason over the topology of graphs while considering those geometric features? Read more
Published:
GraphSAGE is a general inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data Read more
Published:
Thin, fragile local structures, and variable global morphologies, etc., of topological tubular structures like vessels and roads, lead to the difficulty of its accurate segmentation. This paper of DSCNet simultaneously enhance its perception for such tubular structures in 3 stages: feature extraction, feature fusion, and loss constraint. Read more
Published:
总体而言，我们从两个方面改进了之前的Deeplab V3+ & DANN分割模型 Read more
Published:
In the scenario of low-dose CT image reconstruction, supervised deep learning is widely adopted, which generally demands a dataset containing pairs of normal-dose and correspoding low-dose images and is therefore challenging in clinical situations. This paper proposed an unsupervised deep learning method to tackle this problem. Read more
Published:
The origin ensemble (OE) algorithm is a novel statistical method for MMSE reconstruction of emission tomography data. This method allows one to perform reconstruction entirely in the image domain, i.e. without the use of forward and backprojection operations. Read more
Published:
The Radon transform is the transform of our n-dimensional volume to a complete set of (n-1)-dimensional line integrals. Whereas the inverse Radon transform is the transform from our complete (n-1)-dimensional line integrals back to the original image. Read more
Published:
Topology studies properties of spaces that are invariant under any continuous deformation. It is sometimes called “rubber-sheet geometry” because the objects can be stretched and contracted like rubber, but cannot be broken. Read more
Published:
This paper explores the causalities among image features, contexts, and categories to eliminate the biased sobject-context entanglement in the class activation maps thus improving the accuracy of object localization. Read more
Published:
This loss function is a modification of the average Hausdorff distance between two unordered sets of points. The proposed method has no notion of bounding boxes, region proposals, or sliding windows. Read more
Published:
Long time no see. After a tiring semester, I finally have the chance to update posts. Though there is much research progress to be talked about, it demands some more time to collect my thoughts about it. So this post will be a rather simple introduction to a famous equation in both physics and computer graphics, the radiative transfer equation (RTE). Read more
Published:
Integro-Differential Equations, IDEs for short, are equations involving both partial differential and integral operators. They are commonly used in a variety of disciplines of science and engineering topics, while nonlinear IDEs are particularly difficult to solve analytically. Therefore, the idea to incorporate neural networks arises naturally. This post is dedicated to recording my understanding of IDEs and NNs. Read more
Published:
Long time no see. After a tiring semester, I finally have the chance to update posts. Though there is much research progress to be talked about, it demands some more time to collect my thoughts about it. So this post will be a rather simple introduction to a famous equation in both physics and computer graphics, the radiative transfer equation (RTE). Read more
Published:
Conventional trials of physics-informed deep learning (say, PINNs) integrates the dynamics as constraints to guide the approximation of a certain function within. An obvious problem with this idea is that such models cannot be directly applied to different initial and boundary conditions and observations. DeepONet, representing a new trial in physics-informed learning, employed some tricks to get around with this. Read more
Published:
In probability theory, the Gillespie algorithm generates a statistically correct trajectory (possible solution) of a stochastic equation system for which the reaction rates are known. Mathematically, it is a variant of a dynamic Monte Carlo method and similar to the kinetic Monte Carlo methods. Read more
Published:
A general overview of papers read recently, and a reflection on what we have done and what to do next. Read more
Published:
Mainstream AIGC models, like ChatGPT and Stable Diffusion, have been drawing much attention over the recent years. Despite the massive amount of high-quality dataset, advanced network structure and loads of GPUs, one innegligible factor that contributes to their success is the use of RLHF, or reinforcement learning from human feedback. This blog is mostly based on a paper in May 2023 by the Levine Lab at UCB, and probes into how RLHF can be elegantly integrated into diffusion framework. Read more
Published:
Reinforcement learning is a machine learning training method based on rewarding desired behaviors and/or punishing undesired ones. In general, a reinforcement learning agent is able to perceive and interpret its environment, take actions and learn through trial and error. Read more
Published:
总体而言，我们从两个方面改进了之前的Deeplab V3+ & DANN分割模型 Read more
Published:
In probability theory, the Gillespie algorithm generates a statistically correct trajectory (possible solution) of a stochastic equation system for which the reaction rates are known. Mathematically, it is a variant of a dynamic Monte Carlo method and similar to the kinetic Monte Carlo methods. Read more
Published:
Integro-Differential Equations, IDEs for short, are equations involving both partial differential and integral operators. They are commonly used in a variety of disciplines of science and engineering topics, while nonlinear IDEs are particularly difficult to solve analytically. Therefore, the idea to incorporate neural networks arises naturally. This post is dedicated to recording my understanding of IDEs and NNs. Read more
Published:
Generative models, powered by AI, is a big trend now for stuff synthesis (texts, images, and even more specialize usages). This post probes into why and how they are working from a mathematical perspective. Read more
Published:
Reinforcement learning is a machine learning training method based on rewarding desired behaviors and/or punishing undesired ones. In general, a reinforcement learning agent is able to perceive and interpret its environment, take actions and learn through trial and error. Read more
Published:
Topology studies properties of spaces that are invariant under any continuous deformation. It is sometimes called “rubber-sheet geometry” because the objects can be stretched and contracted like rubber, but cannot be broken. Read more
Published:
Conventional trials of physics-informed deep learning (say, PINNs) integrates the dynamics as constraints to guide the approximation of a certain function within. An obvious problem with this idea is that such models cannot be directly applied to different initial and boundary conditions and observations. DeepONet, representing a new trial in physics-informed learning, employed some tricks to get around with this. Read more
Published:
In the scenario of low-dose CT image reconstruction, supervised deep learning is widely adopted, which generally demands a dataset containing pairs of normal-dose and correspoding low-dose images and is therefore challenging in clinical situations. This paper proposed an unsupervised deep learning method to tackle this problem. Read more