Diffusion models have demonstrated significant potential for generating high-quality images, audio, and videos. However, their iterative inference process entails substantial computational costs, limiting practical applications. Recently, researchers have introduced accelerated sampling methods that enable diffusion models to generate samples with far fewer timesteps than those used during training. Nonetheless, as the number of sampling steps decreases, the prediction errors significantly degrade the quality of generated outputs. Additionally, the exposure bias in diffusion models further amplifies these errors. To address these challenges, we leverage a manifold hypothesis to explore the exposure bias problem in depth. Based on this geometric perspective, we propose a manifold constraint that effectively reduces exposure bias during accelerated sampling of diffusion models. Notably, our method involves no additional training and requires only minimal hyperparameter tuning. Extensive experiments demonstrate the effectiveness of our approach, achieving a FID score of 15.60 with 10-step SDXL on MS-COCO, surpassing the baseline by a reduction of 2.57 in FID.

In this paper, we propose a novel task, Universal Category Discovery (UCD), to address the challenge of partial overlap between source and target domain categories. Different from previous tasks that assume all known categories exist in the target domain, UCD introduces "private-known" categories that only exist in the source domain and aims to classify unlabeled data as "common" or "novel" categories while avoiding misclassifying them into "private-known" categories. For this task, we propose a Tri-branch network with bidirectional Prototype-aware Matching (TriPM). TriPM effectively transfers knowledge from labeled to unlabeled data by bidirectionally matching similar data pairs, while a prototype matching strategy reduces the negative transfer risk from "private-known" categories. Finally, we propose a tri-branch network to decouple knowledge acquisition from labeled data, unlabeled data, and their interactions, which can avoid knowledge forgetting, explore novel patterns, and transfer common knowledge, respectively. Experiments demonstrate our model’s superiority over SOTA methods.

In cooperative orthogonal frequency division multiplexing (OFDM) systems, accurate frequency synchronization is critical to achieving any potential gains brought by the cooperative operation. The carrier frequency offsets (CFOs) present among multiple nodes (source, relays and destination) are more difficult to tackle than the single CFO problem in point-to-point systems. Multiple CFOs cause phase drift, inter-carrier interference (ICI) and inter-block interference (IBI) in the received signal. This paper deals with the CFO induced interference mitigation problem in distributed space time block coded (STBC) amplify-and-forward (AF) cooperative OFDM systems. We propose a two step approach to recover the phase distortion and suppress the ICI and IBI using low complexity methods to achieve high performance. The first step is time domain (TD) compensation and the second step is frequency domain (FD) decoding. Two TD compensation schemes are proposed, i.e., IBI-removal and ICI-removal. The IBI-removal scheme decouples the two blocks of one STBC codeword completely and then decodes the ICI degraded blocks individually. The ICI-removal scheme removes ICI first and the subsequent decoding requires joint decoding of the two blocks. Simulation results show that the IBI-removal scheme which is of lower complexity performs well with small CFO. For large CFO, the ICI-removal with modified iterative joint maximum likelihood decoding (MIJMLD) outperforms other schemes.

In this paper, timing synchronization in high-rate ultra-wideband (UWB) block transmission systems is investigated. A new joint timing and channel estimation scheme is proposed for orthogonal frequency division multiplexing (OFDM) and single carrier block transmission with frequency domain equalization (SC-FDE) UWB systems. The scheme is based on a newly designed preamble for both coarse timing and the subsequent channel estimation. Despite of the presence of coarse timing error, the estimated channel impulse response (CIR) is simply the cyclic shifted version of the real CIR, thanks to the unique structure of the preamble. The coarse timing error (CTE) is then determined from the CIR estimation and used to fine tune the timing position and the frequency domain equalization coefficients. The proposed scheme saves preamble overhead by performing joint synchronization and channel estimation, and outperforms existing timing acquisition methods in the literature in dense multipath UWB channels. In addition, the impact of timing error on channel estimation and the performance of SC-FDE UWB systems are analyzed, and the bit-error-rate (BER) degradation with respect to a certain timing error is derived.

This paper deals with the interference mitigation problem of the distributed space time block coded (STBC) amplify-and-forward (AF) cooperative orthogonal frequency division multiplexing (OFDM) signal in the presence of carrier frequency offsets (CFO). Multiple CFOs introduce phase drift, inter-carrier interference (ICI) and inter-block interference (IBI) to the received signal. A joint time domain (TD) and frequency domain (FD) method is presented to recover the phase distortion and mitigate the ICI and IBI with low complexity and high performance. The TD compensation proposed in this paper removes the IBI first and then the ICI is mitigated through FD equalization (FDE) methods. For the FDE, the minimum mean square error (MMSE) equalizer has been derived which has a different noise covariance matrix from the conventional MMSE. Sub-block processing is employed to reduce the computational complexity in FD equalization. Furthermore, a two-pilot-block assisted channel estimation method is proposed for AF cooperative OFDM in the presence of CFOs. The bit-error-rate (BER) performance evaluated via computer simulation demonstrates the effectiveness of the proposed method.

In this paper, a new joint timing and channel estimation scheme is proposed for orthogonal frequency division multiplexing and single carrier block transmission ultra- wideband (UWB) systems. In particular, a new preamble pattern is used for both coarse timing and channel estimation. Despite of the presence of coarse timing error, the estimated channel impulse response (CIR) is simply the cyclic shifted version of the real CIR, thanks to the unique structure of the preamble. Then a new method is proposed to search for the first channel tap in the CIR vector, which can be employed to fine tune the timing position and adjust the channel estimation for later data block detection. The proposed scheme has low complexity, saves preamble overhead by performing joint synchronization and channel estimation, and outperforms existing timing acquisition methods in the literature by a large margin in dense multipath UWB channels. Moreover, this paper presents an analysis of the effect of residual timing error on the performance of single carrier systems with frequency domain equalization.