Search for: All records

Recent approaches have shown promises distilling diffusion models into efficient one-step generators. Among them, Distribution Matching Distillation (DMD) produces one-step generators that match their teacher in distribution, without enforcing a one-to-one correspondence with the sampling trajectories of their teachers. However, to ensure stable training, DMD requires an additional regression loss computed using a large set of noise-image pairs generated by the teacher with many steps of a deterministic sampler. This is costly for large-scale text-to-image synthesis and limits the student's quality, tying it too closely to the teacher's original sampling paths. We introduce DMD2, a set of techniques that lift this limitation and improve DMD training. First, we eliminate the regression loss and the need for expensive dataset construction. We show that the resulting instability is due to the fake critic not estimating the distribution of generated samples accurately and propose a two time-scale update rule as a remedy. Second, we integrate a GAN loss into the distillation procedure, discriminating between generated samples and real images. This lets us train the student model on real data, mitigating the imperfect real score estimation from the teacher model, and enhancing quality. Lastly, we modify the training procedure to enable multi-step sampling. We identify and address the training-inference input mismatch problem in this setting, by simulating inference-time generator samples during training time. Taken together, our improvements set new benchmarks in one-step image generation, with FID scores of 1.28 on ImageNet-64x64 and 8.35 on zero-shot COCO 2014, surpassing the original teacher despite a 500X reduction in inference cost. Further, we show our approach can generate megapixel images by distilling SDXL, demonstrating exceptional visual quality among few-step methods. 

We propose a novel model-based reinforcement learning algorithm—Dynamics Learning and predictive control with Parameterized Actions (DLPA)—for Parameterized Action Markov Decision Processes (PAMDPs). The agent learns a parameterized-action-conditioned dynamics model and plans with a modified Model Predictive Path Integral control. We theoretically quantify the difference between the generated trajectory and the optimal trajectory during planning in terms of the value they achieved through the lens of Lipschitz Continuity. Our empirical results on several standard benchmarks show that our algorithm achieves superior sample efficiency and asymptotic performance than state-of-the-art PAMDP methods. 

An increasing number of location-based service providers are taking the advantage of cloud computing by outsourcing their Point of Interest (POI) datasets and query services to third-party cloud service providers (CSPs), which answer various location-based queries from users on their behalf. A critical security challenge is to ensure the integrity and completeness of any query result returned by CSPs. As an important type of queries, a location-based skyline query (LBSQ) asks for the POIs not dominated by any other POI with respect to a given query position, i.e., no POI is both closer to the query position and more preferable with respect to a given numeric attribute. While there have been several recent attempts on authenticating outsourced LBSQ, none of them support the shortest path distance that is preferable to the Euclidian distance in metropolitan areas. In this paper, we tackle this open challenge by introducing AuthSkySP, a novel scheme for authenticating outsourced LBSQ under the shortest path distance, which allows the user to verify the integrity and completeness of any LBSQ result returned by an untrusted CSP. We confirm the effectiveness and efficiency of our proposed solution via detailed experimental studies using both real and synthetic datasets.