Large Text-to-Image(T2I) diffusion models have shown a remarkable capability to produce photorealistic and diverse images based on text inputs. However, existing works only support limited language input, e.g., English, Chinese, and Japanese, leaving users beyond these languages underserved and blocking the global expansion of T2I models. Therefore, this paper presents AltDiffusion, a novel multilingual T2I diffusion model that supports eighteen different languages. Specifically, we first train a multilingual text encoder based on the knowledge distillation. Then we plug it into a pretrained English-only diffusion model and train the model with a two-stage schema to enhance the multilingual capability, including concept alignment and quality improvement stage on a large-scale multilingual dataset. Furthermore, we introduce a new benchmark, which includes Multilingual-General-18(MG-18) and Multilingual-Cultural-18(MC-18) datasets, to evaluate the capabilities of T2I diffusion models for generating high-quality images and capturing culture-specific concepts in different languages. Experimental results on both MG-18 and MC-18 demonstrate that AltDiffusion outperforms current state-of-the-art T2I models, e.g., Stable Diffusion in multilingual understanding, especially with respect to culture-specific concepts, while still having comparable capability for generating high-quality images. All source code and checkpoints could be found in https://github.com/superhero-7/AltDiffuson.
Occlusion is a common issue in 3D reconstruction from RGB-D videos, often blocking the complete reconstruction of objects and presenting an ongoing problem. In this paper, we propose a novel framework, empowered by a 2D diffusion-based in-painting model, to reconstruct complete surfaces for the hidden parts of objects. Specifically, we utilize a pre-trained diffusion model to fill in the hidden areas of 2D images. Then we use these in-painted images to optimize a neural implicit surface representation for each instance for 3D reconstruction. Since creating the in-painting masks needed for this process is tricky, we adopt a human-in-the-loop strategy that involves very little human engagement to generate high-quality masks. Moreover, some parts of objects can be totally hidden because the videos are usually shot from limited perspectives. To ensure recovering these invisible areas, we develop a cascaded network architecture for predicting signed distance field, making use of different frequency bands of positional encoding and maintaining overall smoothness. Besides the commonly used rendering loss, Eikonal loss, and silhouette loss, we adopt a CLIP-based semantic consistency loss to guide the surface from unseen camera angles. Experiments on ScanNet scenes show that our proposed framework achieves state-of-the-art accuracy and completeness in object-level reconstruction from scene-level RGB-D videos. Code: https://github.com/THU-LYJ-Lab/O2-Recon.
Model Predictive Control (MPC) has proven to be a powerful tool for the control of systems with constraints. Nonetheless, in many applications, a major challenge arises, that is finding the optimal solution within a single sampling instant to apply a receding-horizon policy. In such cases, many suboptimal solutions have been proposed, among which the possibility of "blocking" some moves a-priori. In this paper, we propose a dynamic approach to move blocking, to exploit the solution already available at the previous iteration, and we show not only that such an approach preserves asymptotic stability, but also that the decrease of performance with respect to the ideal solution can be theoretically bounded.
The accuracy of digital elevation models (DEMs) in urban areas is influenced by numerous factors including land cover and terrain irregularities. Moreover, building artifacts in global DEMs cause artificial blocking of surface flow pathways. This compromises their quality and adequacy for hydrological and environmental modelling in urban landscapes where precise and accurate terrain information is needed. In this study, the extreme gradient boosting (XGBoost) ensemble algorithm is adopted for enhancing the accuracy of two medium-resolution 30m DEMs over Cape Town, South Africa: Copernicus GLO-30 and ALOS World 3D (AW3D). XGBoost is a scalable, portable and versatile gradient boosting library that can solve many environmental modelling problems. The training datasets are comprised of eleven predictor variables including elevation, urban footprints, slope, aspect, surface roughness, topographic position index, terrain ruggedness index, terrain surface texture, vector roughness measure, forest cover and bare ground cover. The target variable (elevation error) was calculated with respect to highly accurate airborne LiDAR. After training and testing, the model was applied for correcting the DEMs at two implementation sites. The correction achieved significant accuracy gains which are competitive with other proposed methods. The root mean square error (RMSE) of Copernicus DEM improved by 46 to 53% while the RMSE of AW3D DEM improved by 72 to 73%. These results showcase the potential of gradient boosted trees for enhancing the quality of DEMs, and for improved hydrological modelling in urban catchments.
Multi-object tracking (MOT) at low frame rates can reduce computational, storage and power overhead to better meet the constraints of edge devices. Many existing MOT methods suffer from significant performance degradation in low-frame-rate videos due to significant location and appearance changes between adjacent frames. To this end, we propose to explore collaborative tracking learning (ColTrack) for frame-rate-insensitive MOT in a query-based end-to-end manner. Multiple historical queries of the same target jointly track it with richer temporal descriptions. Meanwhile, we insert an information refinement module between every two temporal blocking decoders to better fuse temporal clues and refine features. Moreover, a tracking object consistency loss is proposed to guide the interaction between historical queries. Extensive experimental results demonstrate that in high-frame-rate videos, ColTrack obtains higher performance than state-of-the-art methods on large-scale datasets Dancetrack and BDD100K, and outperforms the existing end-to-end methods on MOT17. More importantly, ColTrack has a significant advantage over state-of-the-art methods in low-frame-rate videos, which allows it to obtain faster processing speeds by reducing frame-rate requirements while maintaining higher performance. Code will be released at https://github.com/yolomax/ColTrack
Scene graph generation aims to detect visual relationship triplets, (subject, predicate, object). Due to biases in data, current models tend to predict common predicates, e.g. "on" and "at", instead of informative ones, e.g. "standing on" and "looking at". This tendency results in the loss of precise information and overall performance. If a model only uses "stone on road" rather than "stone blocking road" to describe an image, it may be a grave misunderstanding. We argue that this phenomenon is caused by two imbalances: semantic space level imbalance and training sample level imbalance. For this problem, we propose DB-SGG, an effective framework based on debiasing but not the conventional distribution fitting. It integrates two components: Semantic Debiasing (SD) and Balanced Predicate Learning (BPL), for these imbalances. SD utilizes a confusion matrix and a bipartite graph to construct predicate relationships. BPL adopts a random undersampling strategy and an ambiguity removing strategy to focus on informative predicates. Benefiting from the model-agnostic process, our method can be easily applied to SGG models and outperforms Transformer by 136.3%, 119.5%, and 122.6% on mR@20 at three SGG sub-tasks on the SGG-VG dataset. Our method is further verified on another complex SGG dataset (SGG-GQA) and two downstream tasks (sentence-to-graph retrieval and image captioning).
Video Quality Assessment (VQA), which aims to predict the perceptual quality of a video, has attracted raising attention with the rapid development of streaming media technology, such as Facebook, TikTok, Kwai, and so on. Compared with other sequence-based visual tasks (\textit{e.g.,} action recognition), VQA faces two under-estimated challenges unresolved in User Generated Content (UGC) videos. \textit{First}, it is not rare that several frames containing serious distortions (\textit{e.g.,}blocking, blurriness), can determine the perceptual quality of the whole video, while other sequence-based tasks require more frames of equal importance for representations. \textit{Second}, the perceptual quality of a video exhibits a multi-distortion distribution, due to the differences in the duration and probability of occurrence for various distortions. In order to solve the above challenges, we propose \textit{Visual Quality Transformer (VQT)} to extract quality-related sparse features more efficiently. Methodologically, a Sparse Temporal Attention (STA) is proposed to sample keyframes by analyzing the temporal correlation between frames, which reduces the computational complexity from $O(T^2)$ to $O(T \log T)$. Structurally, a Multi-Pathway Temporal Network (MPTN) utilizes multiple STA modules with different degrees of sparsity in parallel, capturing co-existing distortions in a video. Experimentally, VQT demonstrates superior performance than many \textit{state-of-the-art} methods in three public no-reference VQA datasets. Furthermore, VQT shows better performance in four full-reference VQA datasets against widely-adopted industrial algorithms (\textit{i.e.,} VMAF and AVQT).
This paper presents a low-cost network architecture for training large language models (LLMs) at hyperscale. We study the optimal parallelization strategy of LLMs and propose a novel datacenter network design tailored to LLM's unique communication pattern. We show that LLM training generates sparse communication patterns in the network and, therefore, does not require any-to-any full-bisection network to complete efficiently. As a result, our design eliminates the spine layer in traditional GPU clusters. We name this design a Rail-only network and demonstrate that it achieves the same training performance while reducing the network cost by 38% to 77% and network power consumption by 37% to 75% compared to a conventional GPU datacenter. Our architecture also supports Mixture-of-Expert (MoE) models with all-to-all communication through forwarding, with only 8.2% to 11.2% completion time overhead for all-to-all traffic. We study the failure robustness of Rail-only networks and provide insights into the performance impact of different network and training parameters.
In the field of music information retrieval (MIR), cover song identification (CSI) is a challenging task that aims to identify cover versions of a query song from a massive collection. Existing works still suffer from high intra-song variances and inter-song correlations, due to the entangled nature of version-specific and version-invariant factors in their modeling. In this work, we set the goal of disentangling version-specific and version-invariant factors, which could make it easier for the model to learn invariant music representations for unseen query songs. We analyze the CSI task in a disentanglement view with the causal graph technique, and identify the intra-version and inter-version effects biasing the invariant learning. To block these effects, we propose the disentangled music representation learning framework (DisCover) for CSI. DisCover consists of two critical components: (1) Knowledge-guided Disentanglement Module (KDM) and (2) Gradient-based Adversarial Disentanglement Module (GADM), which block intra-version and inter-version biased effects, respectively. KDM minimizes the mutual information between the learned representations and version-variant factors that are identified with prior domain knowledge. GADM identifies version-variant factors by simulating the representation transitions between intra-song versions, and exploits adversarial distillation for effect blocking. Extensive comparisons with best-performing methods and in-depth analysis demonstrate the effectiveness of DisCover and the and necessity of disentanglement for CSI.
Training deep neural networks (DNNs) is computationally intensive but arrays of non-volatile memories like Charge Trap Flash (CTF) can accelerate DNN operations using in-memory computing. Specifically, the Resistive Processing Unit (RPU) architecture uses the voltage-threshold program by stochastic encoded pulse trains and analog memory features to accelerate vector-vector outer product and weight update for the gradient descent algorithms. Although CTF, offering high precision, has been regarded as an excellent choice for implementing RPU, the accumulation of charge due to the applied stochastic pulse trains is ultimately of critical significance in determining the final weight update. In this paper, we report the non-ideal program-time conservation in CTF through pulsing input measurements. We experimentally measure the effect of pulse width and pulse gap, keeping the total ON-time of the input pulse train constant, and report three non-idealities: (1) Cumulative V_T shift reduces when total ON-time is fragmented into a larger number of shorter pulses, (2) Cumulative V_T shift drops abruptly for pulse widths < 2 {\mu}s, (3) Cumulative V_T shift depends on the gap between consecutive pulses and the V_T shift reduction gets recovered for smaller gaps. We present an explanation based on a transient tunneling field enhancement due to blocking oxide trap-charge dynamics to explain these non-idealities. Identifying and modeling the responsible mechanisms and predicting their system-level effects during learning is critical. This non-ideal accumulation is expected to affect algorithms and architectures relying on devices for implementing mathematically equivalent functions for in-memory computing-based acceleration.
The number of train passengers and the demand for high data rates to handle new technologies such as video streaming and IoT technologies are continuously increasing. Therefore the exploration of millimeter waves (mmWave) band is a key technology to meet this demand. However, the high penetration loss makes mmWave very sensitive to blocking, limiting its coverage area. One promising, efficient, and low-cost solution is the reconfigurable intelligent surface (RIS). This paper reviews the state of the art of RIS for railway communications in the mmWave context. First, we present the different types of RIS and review some optimization algorithms used in the literature to find the RIS phase shift. Then, we review recent works on RIS in the railway domain and provide future directions.
In recent years, there is an increasing demand for unmanned aerial vehicles (UAVs) to complete multiple applications. However, as unmanned equipments, UAVs lead to some security risks to general civil aviations. In order to strengthen the flight management of UAVs and guarantee the safety, UAVs can be equipped with automatic dependent surveillance-broadcast (ADS-B) devices. In addition, as an automatic system, ADS-B can periodically broadcast flight information to the nearby aircrafts or the ground stations, and the technology is already used in civil aviation systems. However, due to the limited frequency of ADS-B technique, UAVs equipped with ADS-B devices result in the loss of packets to both UAVs and civil aviation. Further, the operation of civil aviation are seriously interfered. Hence, this paper firstly examines the packets loss of civil planes at different distance, then analyzes the impact of UAVs equipped with ADS-B on the packets updating of civil planes. The result indicates that the 1090MHz band blocking is affected by the density of UAVs. Besides, the frequency capacity is affected by the requirement of updating interval of civil planes. The position updating probability within 3s is 92.3% if there are 200 planes within 50km and 20 UAVs within 5km. The position updating probability within 3s is 86.9% if there are 200 planes within 50km and 40 UAVs within 5km.
The quick rise of emerging technologies like spectrally efficient multicarrier with a higher order modulation and bandwidth variable wavelength selective switches have caused a shift of the optical network architecture from fixed to the flexible elastic optical network (EON), in terms of bit rate, center frequency spacing, modulation format, and optical reach. By using bandwidth-variable transceivers, colorless flexible-grid reconfigurable optical add-drop multiplexers (ROADM), as well as a choice of different optical source types, operators can provide a range of service types and improve network efficiency. Implementation of the EON network requires careful consideration, especially in terms of the ROADM structure to achieve optimal performance. We propose a low-cost ROADM structure with an order-based connection method that includes a high degree with a blocking probability of less than 10-4. Then, we applied the new design of ROADM in EON that transported 1 Tb/s. In the final step, we conclude that our system accommodated 20 percent more traffic demand over an average of the low-rate wavelength approach.
Entity matching (EM) finds data instances that refer to the same real-world entity. Most EM solutions perform blocking then matching. Many works have applied deep learning (DL) to matching, but far fewer works have applied DL to blocking. These blocking works are also limited in that they consider only a simple form of DL and some of them require labeled training data. In this paper, we develop the DeepBlocker framework that significantly advances the state of the art in applying DL to blocking for EM. We first define a large space of DL solutions for blocking, which contains solutions of varying complexity and subsumes most previous works. Next, we develop eight representative solutions in this space. These solutions do not require labeled training data and exploit recent advances in DL (e.g., sequence modeling, transformer, self supervision). We empirically determine which solutions perform best on what kind of datasets (structured, textual, or dirty). We show that the best solutions (among the above eight) outperform the best existing DL solution and the best existing non-DL solutions (including a state-of-the-art industrial non-DL solution), on dirty and textual data, and are comparable on structured data. Finally, we show that the combination of the best DL and non-DL solutions can perform even better, suggesting a new venue for research.
Blocking is a major task in entity matching. Numerous blocking solutions have been developed, but as far as we can tell, blocking using the well-known tf/idf measure has received virtually no attention. Yet, when we experimented with tf/idf blocking using Lucene, we found it did quite well. So in this paper we examine tf/idf blocking in depth. We develop Sparkly, which uses Lucene to perform top-k tf/idf blocking in a distributed share-nothing fashion on a Spark cluster. We develop techniques to identify good attributes and tokenizers that can be used to block on, making Sparkly completely automatic. We perform extensive experiments showing that Sparkly outperforms 8 state-of-the-art blockers. Finally, we provide an in-depth analysis of Sparkly's performance, regarding both recall/output size and runtime. Our findings suggest that (a) tf/idf blocking needs more attention, (b) Sparkly forms a strong baseline that future blocking work should compare against, and (c) future blocking work should seriously consider top-k blocking, which helps improve recall, and a distributed share-nothing architecture, which helps improve scalability, predictability, and extensibility.
We examine here the effects of recurrent vaccination and waning immunity on the establishment of an endemic equilibrium in a population. An individual-based model that incorporates memory effects for transmission rate during infection and subsequent immunity is introduced, considering stochasticity at the individual level. By letting the population size going to infinity, we derive a set of equations describing the large scale behavior of the epidemic. The analysis of the model's equilibria reveals a criterion for the existence of an endemic equilibrium, which depends on the rate of immunity loss and the distribution of time between booster doses. The outcome of a vaccination policy in this context is influenced by the efficiency of the vaccine in blocking transmissions and the distribution pattern of booster doses within the population. Strategies with evenly spaced booster shots at the individual level prove to be more effective in preventing disease spread compared to irregularly spaced boosters, as longer intervals without vaccination increase susceptibility and facilitate more efficient disease transmission. We provide an expression for the critical fraction of the population required to adhere to the vaccination policy in order to eradicate the disease, that resembles a well-known threshold for preventing an outbreak with an imperfect vaccine. We also investigate the consequences of unequal vaccine access in a population and prove that, under reasonable assumptions, fair vaccine allocation is the optimal strategy to prevent endemicity.
When a JPEG image is compressed using the loss compression method with a high compression rate, a blocking phenomenon can occur in the image, making it necessary to restore the image to its original quality. In particular, restoring compressed images that are unrecognizable presents an innovative challenge. Therefore, this paper aims to address the restoration of JPEG images that have suffered significant loss due to maximum compression using a GAN-based net-work method. The generator in this network is based on the U-Net architecture and features a newly presented hourglass structure that can preserve the charac-teristics of deep layers. Additionally, the network incorporates two loss functions, LF Loss and HF Loss, to generate natural and high-performance images. HF Loss uses a pretrained VGG-16 network and is configured using a specific layer that best represents features, which can enhance performance for the high-frequency region. LF Loss, on the other hand, is used to handle the low-frequency region. These two loss functions facilitate the generation of images by the generator that can deceive the discriminator while accurately generating both high and low-frequency regions. The results show that the blocking phe-nomenon in lost compressed images was removed, and recognizable identities were generated. This study represents a significant improvement over previous research in terms of image restoration performance.
Class imbalance is the phenomenon that some classes have much fewer instances than others, which is ubiquitous in real-world graph-structured scenarios. Recent studies find that off-the-shelf Graph Neural Networks (GNNs) would under-represent minor class samples. We investigate this phenomenon and discover that the subspaces of minor classes being squeezed by those of the major ones in the latent space is the main cause of this failure. We are naturally inspired to enlarge the decision boundaries of minor classes and propose a general framework GraphSHA by Synthesizing HArder minor samples. Furthermore, to avoid the enlarged minor boundary violating the subspaces of neighbor classes, we also propose a module called SemiMixup to transmit enlarged boundary information to the interior of the minor classes while blocking information propagation from minor classes to neighbor classes. Empirically, GraphSHA shows its effectiveness in enlarging the decision boundaries of minor classes, as it outperforms various baseline methods in class-imbalanced node classification with different GNN backbone encoders over seven public benchmark datasets. Code is avilable at https://github.com/wenzhilics/GraphSHA.
Recommender systems (RSs) have become an indispensable part of online platforms. With the growing concerns of algorithmic fairness, RSs are not only expected to deliver high-quality personalized content, but are also demanded not to discriminate against users based on their demographic information. However, existing RSs could capture undesirable correlations between sensitive features and observed user behaviors, leading to biased recommendations. Most fair RSs tackle this problem by completely blocking the influences of sensitive features on recommendations. But since sensitive features may also affect user interests in a fair manner (e.g., race on culture-based preferences), indiscriminately eliminating all the influences of sensitive features inevitably degenerate the recommendations quality and necessary diversities. To address this challenge, we propose a path-specific fair RS (PSF-RS) for recommendations. Specifically, we summarize all fair and unfair correlations between sensitive features and observed ratings into two latent proxy mediators, where the concept of path-specific bias (PS-Bias) is defined based on path-specific counterfactual inference. Inspired by Pearl's minimal change principle, we address the PS-Bias by minimally transforming the biased factual world into a hypothetically fair world, where a fair RS model can be learned accordingly by solving a constrained optimization problem. For the technical part, we propose a feasible implementation of PSF-RS, i.e., PSF-VAE, with weakly-supervised variational inference, which robustly infers the latent mediators such that unfairness can be mitigated while necessary recommendation diversities can be maximally preserved simultaneously. Experiments conducted on semi-simulated and real-world datasets demonstrate the effectiveness of PSF-RS.
We investigate the problem of online learning with monotone and continuous DR-submodular reward functions, which has received great attention recently. To efficiently handle this problem, especially in the case with complicated decision sets, previous studies have proposed an efficient projection-free algorithm called Mono-Frank-Wolfe (Mono-FW) using $O(T)$ gradient evaluations and linear optimization steps in total. However, it only attains a $(1-1/e)$-regret bound of $O(T^{4/5})$. In this paper, we propose an improved projection-free algorithm, namely POBGA, which reduces the regret bound to $O(T^{3/4})$ while keeping the same computational complexity as Mono-FW. Instead of modifying Mono-FW, our key idea is to make a novel combination of a projection-based algorithm called online boosting gradient ascent, an infeasible projection technique, and a blocking technique. Furthermore, we consider the decentralized setting and develop a variant of POBGA, which not only reduces the current best regret bound of efficient projection-free algorithms for this setting from $O(T^{4/5})$ to $O(T^{3/4})$, but also reduces the total communication complexity from $O(T)$ to $O(\sqrt{T})$.