Learning low-dimensional semantic representations in diffusion models (DMs) is an open task, since in standard DMs, the dimensions of its intermediate latents are the same as that of the observations and thus are unable to represent low-dimensional semantics. Existing methods address this task either by encoding observations into semantics which makes it difficult to generate samples without observations, or by synthesizing the U-Net's layers of pre-trained DMs into low-dimensional semantics, which is mainly used for downstream tasks rather than using semantics to facilitate the training process. Further, those generated static representations might not be aligned with dynamic timestep-wise intermediate latents. This work introduces a Progressive self-encoded Diffusion model (ProgDiffusion), which simultaneously learns semantic representations and reconstructs observations, does efficient unconditional generation, and produces progressively structured semantic representations. These benefits are gained by a novel self-encoder mechanism which takes the U-Net's upsampling features, intermediate latent and the denoising timestep as conditions to generate time-specific semantic representations, differing from existing work of conditioning on observations only. As a result, the learned intermediate latents are dynamic and mapped to a series of semantic representations that capture their gradual changes. Notably, our proposed encoder operates independently of the observations, making it feasible for unconditional generation as observations are not required. To evaluate ProgDiffusion, we design tasks to visualise the learned progressive semantic representations, in addition to other common tasks, which validate the effectiveness of ProgDiffusion against the state-of-the-art. The code is available at https://github.com/amasawa/ProgDiffusion.
论文检索
输入标题、作者或关键词，从 9,256 篇学术成果中精准定位
Recent advances in neural models have shown considerable promise in solving Traveling Salesman Problems (TSPs) without relying on much hand-crafted engineering. However, while non-autoregressive (NAR) approaches benefit from faster inference through parallelism, they typically deliver solutions of inferior quality compared to autoregressive ones. To enhance the solution quality while maintaining fast inference, we propose DEITSP, a diffusion model with efficient iterations tailored for TSP that operates in a NAR manner. Firstly, we introduce a one-step diffusion model that integrates the controlled discrete noise addition process with self-consistency enhancement, enabling optimal solution prediction through simultaneous denoising of multiple solutions. Secondly, we design a dual-modality graph transformer to bolster the extraction and fusion of features from node and edge modalities, while further accelerating the inference with fewer layers. Thirdly, we develop an efficient iterative strategy that alternates between adding and removing noise to improve exploration compared to previous diffusion methods. Additionally, we devise a scheduling framework to progressively refine the solution space by adjusting noise levels, facilitating a smooth search for optimal solutions. Extensive experiments on real-world and large-scale TSP instances demonstrate that DEITSP performs favorably against existing neural approaches in terms of solution quality, inference latency, and generalization ability.
Computerized Adaptive Testing (CAT) aims to select the most appropriate questions based on the examinee's ability and is widely used in online education. However, existing CAT systems often lack initial understanding of the examinee's ability, requiring random probing questions. This can lead to poorly matched questions, extending the test duration and negatively impacting the examinee's mindset, a phenomenon referred to as the Cold Start with Insufficient Prior (CSIP) task. This issue occurs because CAT systems do not effectively utilize the abundant prior information about the examinee available from other courses on online platforms. These response records, due to the commonality of cognitive states across different knowledge domains, can provide valuable prior information for the target domain. However, no prior work has explored solutions for the CSIP task. In response to this gap, we propose Diffusion Cognitive States TransfeR Framework (DCSR), a novel domain transfer framework based on Diffusion Models (DMs) to address the CSIP task. Specifically, we construct a cognitive state transition bridge between domains, guided by the common cognitive states of examinees, encouraging the model to reconstruct the initial ability state in the target domain. To enrich the expressive power of the generated data, we analyze the causal relationships in the generation process from a causal perspective. Redundant and extraneous cognitive states can lead to limited transfer and negative transfer effects. Therefore, we designed three decoupling strategies to control confounding variables, thereby blocking backdoor paths that hinder causal discovery. Given that excessive uncertainty can affect the applicability of generated results to the CAT system, we propose consistency constraint and task-oriented constraint to control the randomness of the generated results and their relevance to the CAT task, respectively. Our DCSR can seamlessly apply the generated initial ability states in the target domain to existing question selection algorithms, thus improving the cold start performance of the CAT sys- tem. Extensive experiments conducted on five real-world datasets demonstrate that DCSR significantly outperforms existing baseline methods in addressing the CSIP task.
Graph Neural Networks (GNNs) have shown promising results in modeling graphs in various tasks. The training of GNNs, especially on specialized tasks such as bioinformatics, demands extensive expert annotations, which are expensive and usually contain sensitive information of data providers. The trained GNN models are often shared for deployment in the real world. As neural networks can memorize the training samples, the model parameters of GNNs have a high risk of leaking private training data. Our theoretical analysis shows the strong connections between trained GNN parameters and the training graphs used, confirming the training graph leakage issue. However, explorations into training data leakage from trained GNNs are rather limited. Therefore, we investigate a novel problem of stealing graphs from trained GNNs. To obtain high-quality graphs that resemble the target training set, a graph diffusion model with diffusion noise optimization is deployed as a graph generator. Furthermore, we propose a selection method that effectively leverages GNN model parameters to identify training graphs from samples generated by the graph diffusion model. Extensive experiments on real-world datasets demonstrate the effectiveness of the proposed framework in stealing training graphs from the trained GNN.
Cross-domain recommendation (CDR) has been proven as a promising way to alleviate the cold-start issue, in which the most critical problem is how to draw an informative user representation in the target domain via the transfer of user preference existing in the source domain. Prior efforts mostly follow the embedding-and-mapping paradigm, which first integrate the preference into user representation in the source domain, and then perform a mapping function on this representation to the target domain. However, they focus on mapping features across domains, neglecting to explicitly model the preference integration process, which may lead to learning coarse user representation. Diffusion models (DMs), which contribute to more accurate user/item representations due to their explicit information injection capability, have achieved promising performance in recommendation systems. Nevertheless, these DMs-based methods cannot directly account for valuable user preference in other domains, leading to challenges in adapting to the transfer of preference for cold-start users. Consequently, the feasibility of DMs for CDR remains underexplored. To this end, we explore to utilize the explicit information injection capability of DMs for user preference integration and propose a Preference-Guided Diffusion Model for CDR to cold-start users, termed as DMCDR. Specifically, we leverage a preference encoder to establish the preference guidance signal with the user's interaction history in the source domain. Then, we explicitly inject the preference guidance signal into the user representation step by step to guide the reverse process, and ultimately generate the personalized user representation in the target domain, thus achieving the transfer of user preference across domains. Furthermore, we comprehensively explore the impact of six DMs-based variants on CDR. Extensive experiments on three real-world CDR scenarios demonstrate the superiority of our DMCDR over SOTA methods and six DMs-based variants.
As one of the most successful generative models, diffusion models have demonstrated remarkable efficacy in synthesizing high-quality images. These models learn the underlying high-dimensional data distribution in an unsupervised manner. Despite their success, diffusion models are highly data-driven and prone to inheriting the imbalances and biases present in real-world data. Some studies have attempted to address these issues by designing text prompts for known biases or using bias labels to construct unbiased data. While these methods have shown improved results, real-world scenarios often contain various unknown biases, and obtaining bias labels is particularly challenging. In this paper, we emphasize the necessity of mitigating bias in pre-trained diffusion models without relying on auxiliary bias annotations. To tackle this problem, we propose a framework, InvDiff, which aims to learn invariant semantic information for diffusion guidance. Specifically, we propose identifying underlying biases in the training data and designing a novel debiasing training objective. Then, we employ a lightweight trainable module that automatically preserves invariant semantic information and uses it to guide the diffusion model's sampling process toward unbiased outcomes simultaneously. Notably, we only need to learn a small number of parameters in the lightweight learnable module without altering the pre-trained diffusion model. Furthermore, we provide a theoretical guarantee that the implementation of InvDiff is equivalent to reducing the error upper bound of generalization. Extensive experimental results on three publicly available benchmarks demonstrate that InvDiff effectively reduces biases while maintaining the quality of image generation. Our code is available at https://github.com/Hundredl/InvDiff.
Large quantities of social activity data, such as weekly web search volumes and the number of new infections with infectious diseases, reflect peoples' interests and activities. It is important to discover temporal patterns from such data and to forecast future activities accurately. However, modeling and forecasting social activity data streams is difficult because they are high-dimensional and composed of multiple time-varying dynamics such as trends, seasonality, and interest diffusion. In this paper, we propose D-Tracker, a method for continuously capturing time-varying temporal patterns within social activity tensor data streams and forecasting future activities. Our proposed method has the following properties: (a) Interpretable: it incorporates the partial differential equation into a tensor decomposition framework and captures time-varying temporal patterns such as trends, seasonality, and interest diffusion between locations in an interpretable manner; (b) Automatic: it has no hyperparameters and continuously models tensor data streams fully automatically; (c) Scalable: the computation time of D-Tracker is independent of the time series length. Experiments using web search volume data obtained from GoogleTrends, and COVID-19 infection data obtained from COVID-19 Open Data Repository show that our method can achieve higher forecasting accuracy in less computation time than existing methods while extracting the interest diffusion between locations. Our source code and datasets are available at https://github.com/Higashiguchi-Shingo/D-Tracker.
Diversity control is an important task to alleviate bias amplification and filter bubble problems. The desired degree of diversity may fluctuate based on users' daily moods or business strategies. However, existing methods for controlling diversity often lack flexibility, as diversity is decided during training and cannot be easily modified during inference. We propose D3Rec (Disentangled Diffusion model for Diversified Recommendation), an end-to-end method that controls the accuracy-diversity trade-off at inference. D3Rec meets our three desiderata by (1) generating recommendations based on category preferences, (2) controlling category preferences during the inference phase, and (3) adapting to arbitrary targeted category preferences. In the forward process, D3Rec removes category preferences lurking in user interactions by adding noises. Then, in the reverse process, D3Rec generates recommendations through denoising steps while reflecting desired category preferences. Extensive experiments on real-world and synthetic datasets validate the effectiveness of D3Rec in controlling diversity at inference.
MedDiT: A Knowledge-Controlled Diffusion Transformer Framework for Dynamic Medical Image Generation in Virtual Simulated Patient
PDF ↗Medical education relies heavily on Simulated Patients (SPs) to provide a safe environment for students to practice clinical skills, including medical image analysis. However, the high cost of recruiting qualified SPs and the lack of diverse medical imaging datasets have presented significant challenges. To address these issues, this paper introduces MedDiT, a novel knowledge-controlled conversational framework that can dynamically generate plausible medical images aligned with simulated patient symptoms, enabling diverse diagnostic skill training. Specifically, MedDiT integrates various patient Knowledge Graphs (KGs), which describe the attributes and symptoms of patients, to dynamically prompt Large Language Models' (LLMs) behavior and control the patient characteristics, mitigating hallucination during medical conversation. Additionally, a well-tuned Diffusion Transformer (DiT) model is incorporated to generate medical images according to the specified patient attributes in the KG. In this paper, we present the capabilities of MedDiT through a practical demonstration, showcasing its ability to act in diverse simulated patient cases and generate the corresponding medical images. This can provide an abundant and interactive learning experience for students, advancing medical education by offering an immersive simulation platform for future healthcare professionals. The work sheds light on the feasibility of incorporating advanced technologies like LLM, KG, and DiT in education applications, highlighting their potential to address the challenges faced in simulated patient-based medical education.
LivePoem: Improving the Learning Experience of Classical Chinese Poetry with AI-Generated Musical Storyboards
PDF ↗Textbook reading has long dominated classical poetry education in Chinese-speaking communities. However, research has shown that extensive text-based learning can lead to learner disengagement and a less pleasant experience. This paper aims to improve the experience of classical Chinese poetry learning by introducing LivePoem—a system that generates musical storyboards (storyboards with background music) as audiovisual aids to support poetry comprehension. We employ a pre-trained diffusion model for storyboard generation and train a prosody-based poem-to-melody generator using a Transformer model, both validated by standard objective metrics to ensure generation quality. Through a within-subjects study involving 25 non-native Chinese learners, we compared learning outcomes from textbook reading and musical storyboard viewing through standardised reading comprehension tests. Additionally, the learning experience was assessed by the Self-Assessment Manikin (SAM) and an inductive thematic analysis of learners' open-ended feedback. Experimental results show that musical storyboards retained the learning outcomes of textbooks, while more effectively engaging learners and providing a more pleasant learning experience.
Chinese Poetry Intoning, with improvised melodies devoid of fixed musical scores, is crucial for emotional expression and prosodic rendition. However, this cultural heritage faces challenges in propagation due to scant audio records and a scarcity of domain experts. Existing text-to-speech models lack the ability to generate melodious audio, while singing-voice-synthesis models rely on predetermined musical scores, which are all unsuitable for intoning synthesis. Hence, we introduce Chinese Poetry Intoning Synthesis (PIS) as a novel task to reproduce intoning audio and preserve this age-old cultural art. Corresponding to this task, we summarize three-level principles from poetry metrical patterns and construct a diffusion PIS model Intoner based on them. We also collect a multi-style Chinese poetry intoning dataset of text-audio pairs accompanied by feature annotations. Experimental results show that our model effectively learns diverse intoning styles and contents which can synthesize more melodious and vibrant intoning audio. To the best of our knowledge, we are the first to work on poetry intoning synthesis task.
Text-to-image diffusion models can generate high-quality images but lack fine-grained control of visual concepts, limiting their creativity. Thus, we introduce component-controllable personalization, a new task that enables users to customize and reconfigure individual components within concepts. This task faces two challenges: semantic pollution, where undesired elements disrupt the target concept, and semantic imbalance, which causes disproportionate learning of the target concept and component. To address these, we design MagicTailor, a framework that uses Dynamic Masked Degradation to adaptively perturb unwanted visual semantics and Dual-Stream Balancing for more balanced learning of desired visual semantics. The experimental results show that MagicTailor achieves superior performance in this task and enables more personalized and creative image generation.
Symbolic music generation aims to create musical notes, which can help users compose music, such as generating target instrument tracks based on provided source tracks. In practical scenarios where there’s a predefined ensemble of tracks and various composition needs, an efficient and effective generative model that can generate any target tracks based on the other tracks becomes crucial. However, previous efforts have fallen short in addressing this necessity due to limitations in their music representations and models. In this paper, we introduce a framework known as GETMusic, with ``GET'' standing for ``GEnerate music Tracks.'' This framework encompasses a novel music representation ``GETScore'' and a diffusion model ``GETDiff.'' GETScore represents musical notes as tokens and organizes tokens in a 2D structure, with tracks stacked vertically and progressing horizontally over time. At a training step, each track of a music piece is randomly selected as either the target or source. The training involves two processes: In the forward process, target tracks are corrupted by masking their tokens, while source tracks remain as the ground truth; in the denoising process, GETDiff is trained to predict the masked target tokens conditioning on the source tracks. Our proposed representation, coupled with the non-autoregressive generative model, empowers GETMusic to generate music with any arbitrary source-target track combinations.Our experiments demonstrate that the versatile GETMusic outperforms prior works proposed for certain specific composition tasks.
Text-to-music (TTM) generation, which converts textual descriptions into audio, opens up innovative avenues for multimedia creation. Achieving high quality and diversity in this process demands extensive, high-quality data, which are often scarce in available datasets. Most open-source datasets frequently suffer from issues like low-quality waveforms and low text-audio consistency, hindering the advancement of music generation models. To address these challenges, we propose a novel quality-aware training paradigm for generating high-quality, high-musicality music from large-scale, quality-imbalanced datasets. Additionally, by leveraging unique properties in the latent space of musical signals, we adapt and implement a masked diffusion transformer (MDT) model for the TTM task, showcasing its capacity for quality control and enhanced musicality. Furthermore, we introduce a three-stage caption refinement approach to address low-quality captions' issue. Experiments show state-of-the-art (SOTA) performance on benchmark datasets including MusicCaps and the Song-Describer Dataset with both objective and subjective metrics. Demo audio samples are available at https://qa-mdt.github.io/, code and pretrained checkpoints are open-sourced at https://github.com/ivcylc/OpenMusic.
SmartSpatial: Enhancing 3D Spatial Awareness in Stable Diffusion with a Novel Evaluation Framework
PDF ↗Stable Diffusion models have made remarkable strides in generating photorealistic images from text prompts but often falter when tasked with accurately representing complex spatial arrangements, particularly involving intricate 3D relationships. To address this limitation, we introduce SmartSpatial, an innovative approach that not only enhances the spatial arrangement capabilities of Stable Diffusion but also fosters AI-assisted creative workflows through 3D-aware conditioning and attention-guided mechanisms. SmartSpatial incorporates depth information injection and cross-attention control to ensure precise object placement, delivering notable improvements in spatial accuracy metrics. In conjunction with SmartSpatial, we present SmartSpatialEval, a comprehensive evaluation framework that bridges computational spatial accuracy with qualitative artistic assessments. Experimental results show that SmartSpatial significantly outperforms existing methods, setting new benchmarks for spatial fidelity in AI-driven art and creativity.
With the emergence of diffusion models and the rapid development of image processing, generating artistic images in style transfer tasks has become effortless. However, these impressive image processing approaches face consistency issues in video processing due to the independent processing of each frame. In this paper, we propose a powerful, model-free approach called FastBlend to address the consistency problem in video stylization. FastBlend functions as a post-processor and can be seamlessly integrated with diffusion models to create a robust video stylization pipeline. Based on a patch-matching algorithm, we remap and blend the aligned content across multiple frames, thus compensating for inconsistent content with neighboring frames. Moreover, we propose a tree-like data structure and a specialized loss function, aiming to optimize computational efficiency and visual quality for different application scenarios. Extensive experiments have demonstrated the effectiveness of FastBlend. Compared with both independent video deflickering algorithms and diffusion-based video processing methods, FastBlend is capable of synthesizing more coherent and realistic videos.
Recently, advancements in video synthesis have attracted significant attention. Video synthesis models have demonstrated the practical applicability of diffusion models in creating dynamic visual content. Despite these advancements, the extension of video lengths remains constrained by computational resources. Most existing video synthesis models are limited to generating short video clips. In this paper, we propose a novel post-tuning methodology for video synthesis models, called ExVideo. This approach is designed to enhance the capability of current video synthesis models, allowing them to produce content over extended temporal durations while incurring lower training expenditures. In particular, we design extension strategies across common temporal model architectures respectively, including 3D convolution, temporal attention, and positional embedding. To evaluate the efficacy of our proposed post-tuning approach, we trained ExSVD, an extended model based on Stable Video Diffusion model. Our approach enhances the model's capacity to generate up to 5x its original number of frames, requiring only 1.5k GPU hours of training on a dataset comprising 40k videos. Importantly, the substantial increase in video length doesn't compromise the model's innate generalization capabilities, and the model showcases its advantages in generating videos of diverse styles and resolutions. We have released the source code and the enhanced model publicly.
CoDiCast: Conditional Diffusion Model for Global Weather Forecasting with Uncertainty Quantification
PDF ↗Accurate weather forecasting is critical for science and society. However, existing methods have not achieved the combination of high accuracy, low uncertainty, and high computational efficiency simultaneously. On one hand, traditional numerical weather prediction (NWP) models are computationally intensive because of their complexity. On the other hand, most machine learning-based weather prediction (MLWP) approaches offer efficiency and accuracy but remain deterministic, lacking the ability to capture forecast uncertainty. To tackle these challenges, we propose a conditional diffusion model, CoDiCast, to generate global weather prediction, integrating accuracy and uncertainty quantification at a modest computational cost. The key idea behind the prediction task is to generate realistic weather scenarios at a future time point, conditioned on observations from the recent past. Due to the probabilistic nature of diffusion models, they can be properly applied to capture the uncertainty of weather predictions. Therefore, we accomplish uncertainty quantifications by repeatedly sampling from stochastic Gaussian noise for each initial weather state and running the denoising process multiple times. Experimental results demonstrate that CoDiCast outperforms several existing MLWP methods in accuracy, and is faster than NWP models in inference speed. Our model can generate 6-day global weather forecasts, at 6-hour steps and 5.625-degree latitude-longitude resolutions, for over 5 variables, in about 12 minutes on a commodity A100 GPU machine with 80GB memory. The source code is available at https://github.com/JimengShi/CoDiCast.
Heatwaves pose a significant threat to public health, especially as global warming intensifies. However, current routing systems (e.g., online maps) fail to incorporate shade information due to the difficulty of estimating shades directly from noisy satellite imagery and the limited availability of training data for generative models. In this paper, we address these challenges through two main contributions. First, we build an extensive dataset covering diverse longitude-latitude regions, varying levels of building density, and different urban layouts. Leveraging Blender-based 3D simulations alongside building outlines, we capture building shadows under various solar zenith angles throughout the year and at different times of day. These simulated shadows are aligned with satellite images, providing a rich resource for learning shade patterns. Second, we propose the DeepShade, a diffusion-based model designed to learn and synthesize shade variations over time. It emphasizes the nuance of edge features by jointly considering RGB with the Canny edge layer, and incorporates contrastive learning to capture the temporal change rules of shade. Then, by conditioning on textual descriptions of known conditions (e.g., time of day, solar angles), our framework provides improved performance in generating shade images. We demonstrate the utility of our approach by using our shade predictions to calculate shade ratios for real-world route planning in Tempe, Arizona. We believe this work will benefit society by providing a reference for urban planning in extreme heat weather and its potential practical applications in the environment.
Map2Traj: Street Map Piloted Zero-shot Trajectory Generation Method for Wireless Network Optimization
PDF ↗In modern wireless networks, user mobility modeling plays a pivotal role in learning-based network optimization, particularly in tasks such as user association and resource allocation. Traditional random mobility models, e.g., random waypoint and Gauss Markov model, often fail to accurately capture the distribution patterns of users within real-world areas. While trace-based mobility models and advanced learning-based trajectory generation methods offer improvements, they are frequently limited by the scarcity of real-world trajectory data in target areas, primarily due to privacy concerns. This paper introduces Map2Traj, a novel zero-shot trajectory generation method that leverages the diffusion model to capture the intrinsic relationship between street maps and user mobility. With solely the street map of an unobserved area, Map2Traj generates synthetic user trajectories that closely resemble the real-world ones in trajectory pattern and spatial distribution. This enables the creation of high-fidelity individual user channel states and an accurate representation of the overall network user distribution, facilitating effective wireless network optimization. Extensive experiments across multiple regions in Xi'an and Chengdu, China demonstrate the effectiveness of our proposed method for zero-shot trajectory generation. A case study applying Map2Traj to user association and load balancing in wireless networks is also presented to validate its efficacy in network optimization.