Gang Chen

Gang Chen

Diffusion probabilistic models have generated high quality image synthesis recently. However, one pain point is the notorious inference to gradually obtain clear images with thousands of steps, which is time consuming compared to other generative models. In this paper, we present a shortcut MCMC sampling algorithm, which balances training and inference, while keeping the generated data's quality. In particular, we add the global fidelity constraint with shortcut MCMC sampling to combat the local fitting from diffusion models. We do some initial experiments and show very promising results. Our implementation is available at https://github.com//vividitytech/diffusion-mcmc.git.

Ran Wang, Xiaoxuan Liu, Hao Ren et al. · tsinghua

Structured decoding enables large language models (LLMs) to generate outputs in formats required by downstream systems, such as HTML or JSON. However, existing methods suffer from efficiency bottlenecks due to grammar compilation, state tracking, and mask creation. We observe that many real-world tasks embed strong prior knowledge about output structure. Leveraging this, we propose a decomposition of constraints into static and dynamic components -- precompiling static structures offline and instantiating dynamic arguments at runtime using grammar snippets. Instead of relying on pushdown automata, we employ a compositional set of operators to model regular formats, achieving lower transition latency. We introduce wgrammar, a lightweight decoding engine that integrates domain-aware simplification, constraint decomposition, and mask caching, achieving up to 250x speedup over existing systems. wgrammar's source code is publicly available at https://github.com/wrran/wgrammar.

Gang Chen

We present a general and simple text to video model based on Transformer. Since both text and video are sequential data, we encode both texts and images into the same hidden space, which are further fed into Transformer to capture the temporal consistency and then decoder to generate either text or images. Considering the image signal may become weak in the long sequence, we introduce the U-Net to reconstruct image from its noised version. Specifically, we increase the noise level to the original image in the long sequence, then use the $down$ module from U-Net to encode noised images, which are further input to transformer to predict next clear images. We also add a constraint to promote motion between any generated image pair in the video. We use GPT2 and test our approach on UCF101 dataset and show it can generate promising videos.

Shaojin Cai, Yuyang Xue3 Qinquan Gao, Min Du et al.

Digitized pathological diagnosis has been in increasing demand recently. It is well known that color information is critical to the automatic and visual analysis of pathological slides. However, the color variations due to various factors not only have negative impact on pathologist's diagnosis, but also will reduce the robustness of the algorithms. The factors that cause the color differences are not only in the process of making the slices, but also in the process of digitization. Different strategies have been proposed to alleviate the color variations. Most of such techniques rely on collecting color statistics to perform color matching across images and highly dependent on a reference template slide. Since the pathological slides between hospitals are usually unpaired, these methods do not yield good matching results. In this work, we propose a novel network that we refer to as Transitive Adversarial Networks (TAN) to transfer the color information among slides from different hospitals or centers. It is not necessary for an expert to pick a representative reference slide in the proposed TAN method. We compare the proposed method with the state-of-the-art methods quantitatively and qualitatively. Compared with the state-of-the-art methods, our method yields an improvement of 0.87dB in terms of PSNR, demonstrating the effectiveness of the proposed TAN method in stain style transfer.

Gang Chen, Sargur N. Srihari

Multimodal learning with deep Boltzmann machines (DBMs) is an generative approach to fuse multimodal inputs, and can learn the shared representation via Contrastive Divergence (CD) for classification and information retrieval tasks. However, it is a 2-fan DBM model, and cannot effectively handle multiple prediction tasks. Moreover, this model cannot recover the hidden representations well by sampling from the conditional distribution when more than one modalities are missing. In this paper, we propose a K-fan deep structure model, which can handle the multi-input and muti-output learning problems effectively. In particular, the deep structure has K-branch for different inputs where each branch can be composed of a multi-layer deep model, and a shared representation is learned in an discriminative manner to tackle multimodal tasks. Given the deep structure, we propose two objective functions to handle two multi-input and multi-output tasks: joint visual restoration and labeling, and the multi-view multi-calss object recognition tasks. To estimate the model parameters, we initialize the deep model parameters with CD to maximize the joint distribution, and then we use backpropagation to update the model according to specific objective function. The experimental results demonstrate that the model can effectively leverages multi-source information and predict multiple tasks well over competitive baselines.

Gang Chen

Semi-supervised clustering is an very important topic in machine learning and computer vision. The key challenge of this problem is how to learn a metric, such that the instances sharing the same label are more likely close to each other on the embedded space. However, little attention has been paid to learn better representations when the data lie on non-linear manifold. Fortunately, deep learning has led to great success on feature learning recently. Inspired by the advances of deep learning, we propose a deep transductive semi-supervised maximum margin clustering approach. More specifically, given pairwise constraints, we exploit both labeled and unlabeled data to learn a non-linear mapping under maximum margin framework for clustering analysis. Thus, our model unifies transductive learning, feature learning and maximum margin techniques in the semi-supervised clustering framework. We pretrain the deep network structure with restricted Boltzmann machines (RBMs) layer by layer greedily, and optimize our objective function with gradient descent. By checking the most violated constraints, our approach updates the model parameters through error backpropagation, in which deep features are learned automatically. The experimental results shows that our model is significantly better than the state of the art on semi-supervised clustering.