Guided diffusion-model generation is a promising direction for customizing the generation process of a pre-trained diffusion-model to address the specific downstream tasks. Existing guided diffusion models either rely on training of the guidance model with pre-collected datasets or require the objective functions to be differentiable. However, for most real-world tasks, the offline datasets are often unavailable, and their objective functions are often not differentiable, such as image generation with human preferences, molecular generation for drug discovery, and material design. Thus, we need an $\textbf{online}$ algorithm capable of collecting data during runtime and supporting a $\textbf{black-box}$ objective function. Moreover, the $\textbf{query efficiency}$ of the algorithm is also critical because the objective evaluation of the query is often expensive in the real-world scenarios. In this work, we propose a novel and simple algorithm, $\textbf{Fast Direct}$, for query-efficient online black-box target generation. Our Fast Direct builds a pseudo-target on the data manifold to update the noise sequence of the diffusion model with a universal direction, which is promising to perform query-efficient guided generation. Extensive experiments on twelve high-resolution ($\small {1024 \times 1024}$) image target generation tasks and six 3D-molecule target generation tasks show $\textbf{6}\times$ up to $\textbf{10}\times$ query efficiency improvement and $\textbf{11}\times$ up to $\textbf{44}\times$ query efficiency improvement, respectively. Our implementation is publicly available at: https://github.com/kimyong95/guide-stable-diffusion/tree/fast-direct

Learning diverse and high-performance behaviors from a limited set of demonstrations is a grand challenge. Traditional imitation learning methods usually fail in this task because most of them are designed to learn one specific behavior even with multiple demonstrations. Therefore, novel techniques for quality diversity imitation learning are needed to solve the above challenge. This work introduces Wasserstein Quality Diversity Imitation Learning (WQDIL), which 1) improves the stability of imitation learning in the quality diversity setting with latent adversarial training based on a Wasserstein Auto-Encoder (WAE), and 2) mitigates a behavior-overfitting issue using a measure-conditioned reward function with a single-step archive exploration bonus. Empirically, our method significantly outperforms state-of-the-art IL methods, achieving near-expert or beyond-expert QD performance on the challenging continuous control tasks derived from MuJoCo environments.

Black-box optimization algorithms have been widely used in various machine learning problems, including reinforcement learning and prompt fine-tuning. However, directly optimizing the training loss value, as commonly done in existing black-box optimization methods, could lead to suboptimal model quality and generalization performance. To address those problems in black-box optimization, we propose a novel Sharpness-Aware Black-box Optimization (SABO) algorithm, which applies a sharpness-aware minimization strategy to improve the model generalization. Specifically, the proposed SABO method first reparameterizes the objective function by its expectation over a Gaussian distribution. Then it iteratively updates the parameterized distribution by approximated stochastic gradients of the maximum objective value within a small neighborhood around the current solution in the Gaussian distribution space. Theoretically, we prove the convergence rate and generalization bound of the proposed SABO algorithm. Empirically, extensive experiments on the black-box prompt fine-tuning tasks demonstrate the effectiveness of the proposed SABO method in improving model generalization performance.

Imitation learning (IL) has shown great potential in various applications, such as robot control. However, traditional IL methods are usually designed to learn only one specific type of behavior since demonstrations typically correspond to a single expert. In this work, we introduce the first generic framework for Quality Diversity Imitation Learning (QD-IL), which enables the agent to learn a broad range of skills from limited demonstrations. Our framework integrates the principles of quality diversity with adversarial imitation learning (AIL) methods, and can potentially improve any inverse reinforcement learning (IRL) method. Empirically, our framework significantly improves the QD performance of GAIL and VAIL on the challenging continuous control tasks derived from Mujoco environments. Moreover, our method even achieves 2x expert performance in the most challenging Humanoid environment.

Diffusion models have demonstrated great potential in generating high-quality content for images, natural language, protein domains, etc. However, how to perform user-preferred targeted generation via diffusion models with only black-box target scores of users remains challenging. To address this issue, we first formulate the fine-tuning of the targeted reserve-time stochastic differential equation (SDE) associated with a pre-trained diffusion model as a sequential black-box optimization problem. Furthermore, we propose a novel covariance-adaptive sequential optimization algorithm to optimize cumulative black-box scores under unknown transition dynamics. Theoretically, we prove a $O(\frac{d^2}{\sqrt{T}})$ convergence rate for cumulative convex functions without smooth and strongly convex assumptions. Empirically, experiments on both numerical test problems and target-guided 3D-molecule generation tasks show the superior performance of our method in achieving better target scores.

The remarkable success of modern machine learning models on large datasets often demands extensive training time and resource consumption. To save cost, a prevalent research line, known as online batch selection, explores selecting informative subsets during the training process. Although recent efforts achieve advancements by measuring the impact of each sample on generalization, their reliance on additional reference models inherently limits their practical applications, when there are no such ideal models available. On the other hand, the vanilla reference-model-free methods involve independently scoring and selecting data in a sample-wise manner, which sacrifices the diversity and induces the redundancy. To tackle this dilemma, we propose Diversified Batch Selection (DivBS), which is reference-model-free and can efficiently select diverse and representative samples. Specifically, we define a novel selection objective that measures the group-wise orthogonalized representativeness to combat the redundancy issue of previous sample-wise criteria, and provide a principled selection-efficient realization. Extensive experiments across various tasks demonstrate the significant superiority of DivBS in the performance-speedup trade-off. The code is publicly available.

Text-guided image editing faces significant challenges to training and inference flexibility. Much literature collects large amounts of annotated image-text pairs to train text-conditioned generative models from scratch, which is expensive and not efficient. After that, some approaches that leverage pre-trained vision-language models are put forward to avoid data collection, but they are also limited by either per text-prompt optimization or inference-time hyper-parameters tuning. To address these issues, we investigate and identify a specific space, referred to as CLIP DeltaSpace, where the CLIP visual feature difference of two images is semantically aligned with the CLIP textual feature difference of their corresponding text descriptions. Based on DeltaSpace, we propose a novel framework called DeltaEdit, which maps the CLIP visual feature differences to the latent space directions of a generative model during the training phase, and predicts the latent space directions from the CLIP textual feature differences during the inference phase. And this design endows DeltaEdit with two advantages: (1) text-free training; (2) generalization to various text prompts for zero-shot inference. Extensive experiments validate the effectiveness and versatility of DeltaEdit with different generative models, including both the GAN model and the diffusion model, in achieving flexible text-guided image editing. Code is available at https://github.com/Yueming6568/DeltaEdit.

Many machine learning applications encounter a situation where model providers are required to further refine the previously trained model so as to gratify the specific need of local users. This problem is reduced to the standard model tuning paradigm if the target data is permissibly fed to the model. However, it is rather difficult in a wide range of practical cases where target data is not shared with model providers but commonly some evaluations about the model are accessible. In this paper, we formally set up a challenge named \emph{Earning eXtra PerformancE from restriCTive feEDdbacks} (EXPECTED) to describe this form of model tuning problems. Concretely, EXPECTED admits a model provider to access the operational performance of the candidate model multiple times via feedback from a local user (or a group of users). The goal of the model provider is to eventually deliver a satisfactory model to the local user(s) by utilizing the feedbacks. Unlike existing model tuning methods where the target data is always ready for calculating model gradients, the model providers in EXPECTED only see some feedbacks which could be as simple as scalars, such as inference accuracy or usage rate. To enable tuning in this restrictive circumstance, we propose to characterize the geometry of the model performance with regard to model parameters through exploring the parameters' distribution. In particular, for the deep models whose parameters distribute across multiple layers, a more query-efficient algorithm is further tailor-designed that conducts layerwise tuning with more attention to those layers which pay off better. Extensive experiments on different applications demonstrate that our work forges a sound solution to the EXPECTED problem. Code is available via https://github.com/kylejingli/EXPECTED.

To ensure that the data collected from human subjects is entrusted with a secret, rival labels are introduced to conceal the information provided by the participants on purpose. The corresponding learning task can be formulated as a noisy partial-label learning problem. However, conventional partial-label learning (PLL) methods are still vulnerable to the high ratio of noisy partial labels, especially in a large labelling space. To learn a more robust model, we present Adversary-Aware Partial Label Learning and introduce the $\textit{rival}$, a set of noisy labels, to the collection of candidate labels for each instance. By introducing the rival label, the predictive distribution of PLL is factorised such that a handy predictive label is achieved with less uncertainty coming from the transition matrix, assuming the rival generation process is known. Nonetheless, the predictive accuracy is still insufficient to produce an sufficiently accurate positive sample set to leverage the clustering effect of the contrastive loss function. Moreover, the inclusion of rivals also brings an inconsistency issue for the classifier and risk function due to the intractability of the transition matrix. Consequently, an adversarial teacher within momentum (ATM) disambiguation algorithm is proposed to cope with the situation, allowing us to obtain a provably consistent classifier and risk function. In addition, our method has shown high resiliency to the choice of the label noise transition matrix. Extensive experiments demonstrate that our method achieves promising results on the CIFAR10, CIFAR100 and CUB200 datasets.

Recent studies show a close connection between neural networks (NN) and kernel methods. However, most of these analyses (e.g., NTK) focus on the influence of (infinite) width instead of the depth of NN models. There remains a gap between theory and practical network designs that benefit from the depth. This paper first proposes a novel kernel family named Neural Optimization Kernel (NOK). Our kernel is defined as the inner product between two $T$-step updated functionals in RKHS w.r.t. a regularized optimization problem. Theoretically, we proved the monotonic descent property of our update rule for both convex and non-convex problems, and a $O(1/T)$ convergence rate of our updates for convex problems. Moreover, we propose a data-dependent structured approximation of our NOK, which builds the connection between training deep NNs and kernel methods associated with NOK. The resultant computational graph is a ResNet-type finite width NN. Our structured approximation preserved the monotonic descent property and $O(1/T)$ convergence rate. Namely, a $T$-layer NN performs $T$-step monotonic descent updates. Notably, we show our $T$-layered structured NN with ReLU maintains a $O(1/T)$ convergence rate w.r.t. a convex regularized problem, which explains the success of ReLU on training deep NN from a NN architecture optimization perspective. For the unsupervised learning and the shared parameter case, we show the equivalence of training structured NN with GD and performing functional gradient descent in RKHS associated with a fixed (data-dependent) NOK at an infinity-width regime. For finite NOKs, we prove generalization bounds. Remarkably, we show that overparameterized deep NN (NOK) can increase the expressive power to reduce empirical risk and reduce the generalization bound at the same time. Extensive experiments verify the robustness of our structured NOK blocks.