Collaborative robots (cobots) are increasingly used to help human workers perform assembly tasks or complete assembly tasks themselves in assembly lines. The ergonomic risks of human workers are a key factor influencing assembly line efficiency. Therefore, this study investigates the mixed-product-model assembly line balancing and sequencing problem (ALBSP) with cobots, considering ergonomic risks in cases where human workers and cobots can operate different tasks in parallel. A mixed-integer programming model is formulated to optimize the makespan and ergonomic risks; this model can solve small-scale instances optimally using the CPLEX solver. A Q-learning-based multiobjective coevolutionary algorithm (QMOCEA) is then developed to handle large-scale instances. This algorithm adopts five vectors for encoding: the task assignment vector handles the task allocation subproblem, the worker allocation vector handles the worker allocation subproblem, the cobot allocation vector handles the cobot allocation subproblem, the process alternative selection vector handles the process alternative selection subproblem, and the product model sequencing vector handles the product model sequencing subproblem. Additionally, this algorithm uses knowledge-based decoding and initialization to obtain high-quality initial solutions. A parameter self-update strategy is proposed to adjust algorithm parameters dynamically. Comparative analysis demonstrates that the proposed method outperforms the original version and exhibits promising performance in comparison with benchmark methods, achieving the highest average hypervolume (HV) ratio of 0.805 and the lowest inverted generational distance (IGD) of 0.028 across 22 instance groups.

The success of large generative models has driven a paradigm shift, leveraging massive multi-source data to enhance model capabilities. However, the interaction among these sources remains theoretically underexplored. This paper provides a first attempt to fill this gap by rigorously analyzing multi-source training in conditional generative modeling, where each condition represents a distinct data source. Specifically, we establish a general distribution estimation error bound in average total variation distance for conditional maximum likelihood estimation based on the bracketing number. Our result shows that when source distributions share certain similarities and the model is expressive enough, multi-source training guarantees a sharper bound than single-source training. We further instantiate the general theory on conditional Gaussian estimation and deep generative models including autoregressive and flexible energy-based models, by characterizing their bracketing numbers. The results highlight that the number of sources and similarity among source distributions improve the advantage of multi-source training. Simulations and real-world experiments validate our theory.

We present a unified probabilistic formulation for diffusion-based image editing, where a latent variable is edited in a task-specific manner and generally deviates from the corresponding marginal distribution induced by the original stochastic or ordinary differential equation (SDE or ODE). Instead, it defines a corresponding SDE or ODE for editing. In the formulation, we prove that the Kullback-Leibler divergence between the marginal distributions of the two SDEs gradually decreases while that for the ODEs remains as the time approaches zero, which shows the promise of SDE in image editing. Inspired by it, we provide the SDE counterparts for widely used ODE baselines in various tasks including inpainting and image-to-image translation, where SDE shows a consistent and substantial improvement. Moreover, we propose \emph{SDE-Drag} -- a simple yet effective method built upon the SDE formulation for point-based content dragging. We build a challenging benchmark (termed \emph{DragBench}) with open-set natural, art, and AI-generated images for evaluation. A user study on DragBench indicates that SDE-Drag significantly outperforms our ODE baseline, existing diffusion-based methods, and the renowned DragGAN. Our results demonstrate the superiority and versatility of SDE in image editing and push the boundary of diffusion-based editing methods. See the project page \url{https://ml-gsai.github.io/SDE-Drag-demo/} for the code and DragBench dataset.

Gradient-based bilevel programming leverages unrolling differentiation (UD) or implicit function theorem (IFT) to solve hyperparameter optimization (HO) problems, and is proven effective and scalable in practice. To understand their generalization behavior, existing works establish upper bounds on the uniform stability of these algorithms, while their tightness is still unclear. To this end, this paper attempts to establish stability lower bounds for UD-based and IFT-based algorithms. A central technical challenge arises from the dependency of each outer-level update on the concurrent stage of inner optimization in bilevel programming. To address this problem, we introduce lower-bounded expansion properties to characterize the instability in update rules which can serve as general tools for lower-bound analysis. These properties guarantee the hyperparameter divergence at the outer level and the Lipschitz constant of inner output at the inner level in the context of HO. Guided by these insights, we construct a quadratic example that yields tight lower bounds for the UD-based algorithm and meaningful bounds for a representative IFT-based algorithm. Our tight result indicates that uniform stability has reached its limit in stability analysis for the UD-based algorithm.

Autoregressively trained transformers have brought a profound revolution to the world, especially with their in-context learning (ICL) ability to address downstream tasks. Recently, several studies suggest that transformers learn a mesa-optimizer during autoregressive (AR) pretraining to implement ICL. Namely, the forward pass of the trained transformer is equivalent to optimizing an inner objective function in-context. However, whether the practical non-convex training dynamics will converge to the ideal mesa-optimizer is still unclear. Towards filling this gap, we investigate the non-convex dynamics of a one-layer linear causal self-attention model autoregressively trained by gradient flow, where the sequences are generated by an AR process $x_{t+1} = W x_t$. First, under a certain condition of data distribution, we prove that an autoregressively trained transformer learns $W$ by implementing one step of gradient descent to minimize an ordinary least squares (OLS) problem in-context. It then applies the learned $\widehat{W}$ for next-token prediction, thereby verifying the mesa-optimization hypothesis. Next, under the same data conditions, we explore the capability limitations of the obtained mesa-optimizer. We show that a stronger assumption related to the moments of data is the sufficient and necessary condition that the learned mesa-optimizer recovers the distribution. Besides, we conduct exploratory analyses beyond the first data condition and prove that generally, the trained transformer will not perform vanilla gradient descent for the OLS problem. Finally, our simulation results verify the theoretical results.

A large-scale deep model pre-trained on massive labeled or unlabeled data transfers well to downstream tasks. Linear evaluation freezes parameters in the pre-trained model and trains a linear classifier separately, which is efficient and attractive for transfer. However, little work has investigated the classifier in linear evaluation except for the default logistic regression. Inspired by the statistical efficiency of naive Bayes, the paper revisits the classical topic on discriminative vs. generative classifiers. Theoretically, the paper considers the surrogate loss instead of the zero-one loss in analyses and generalizes the classical results from binary cases to multiclass ones. We show that, under mild assumptions, multiclass naive Bayes requires $O(\log n)$ samples to approach its asymptotic error while the corresponding multiclass logistic regression requires $O(n)$ samples, where $n$ is the feature dimension. To establish it, we present a multiclass $\mathcal{H}$-consistency bound framework and an explicit bound for logistic loss, which are of independent interests. Simulation results on a mixture of Gaussian validate our theoretical findings. Experiments on various pre-trained deep vision models show that naive Bayes consistently converges faster as the number of data increases. Besides, naive Bayes shows promise in few-shot cases and we observe the "two regimes” phenomenon in pre-trained supervised models. Our code is available at https://github.com/ML-GSAI/Revisiting-Dis-vs-Gen-Classifiers.

Generative data augmentation, which scales datasets by obtaining fake labeled examples from a trained conditional generative model, boosts classification performance in various learning tasks including (semi-)supervised learning, few-shot learning, and adversarially robust learning. However, little work has theoretically investigated the effect of generative data augmentation. To fill this gap, we establish a general stability bound in this not independently and identically distributed (non-i.i.d.) setting, where the learned distribution is dependent on the original train set and generally not the same as the true distribution. Our theoretical result includes the divergence between the learned distribution and the true distribution. It shows that generative data augmentation can enjoy a faster learning rate when the order of divergence term is $o(\max\left( \log(m)\beta_m, 1 / \sqrt{m})\right)$, where $m$ is the train set size and $\beta_m$ is the corresponding stability constant. We further specify the learning setup to the Gaussian mixture model and generative adversarial nets. We prove that in both cases, though generative data augmentation does not enjoy a faster learning rate, it can improve the learning guarantees at a constant level when the train set is small, which is significant when the awful overfitting occurs. Simulation results on the Gaussian mixture model and empirical results on generative adversarial nets support our theoretical conclusions.

Unsupervised domain adaptive (UDA) land-cover classification has recently gained more and more attention. UDA aimed to learn a model from the annotated source data and the unlabeled target data that can perform well on the target domain. The existing UDA frameworks based on adversarial training and self-training methods have boosted this field a lot. However, these methods almost all originate from the computer vision field, and they ignore the very nature of high-resolution remote sensing (HRS) images. The core insight of this paper is that a good land-cover classification result always has strong local consistency and good global diversity, which makes it possible to construct a metric representing the properties of good land-cover mapping, to improve the existing UDA algorithms. Firstly, based on this finding, we prove that local consistency and global diversity can be measured by the Frobenius norm and nuclear norm, respectively. Secondly, we propose a novel local consistency and global diversity metric (LCGDM), which can be easily integrated into the existing UDA frameworks. Finally, the experiments conducted on the LoveDA data set prove the validity of the proposed metric, which can not only improve the overall land-cover mapping but also the category-wise prediction.

Land-cover classification has long been a hot and difficult challenge in remote sensing community. With massive High-resolution Remote Sensing (HRS) images available, manually and automatically designed Convolutional Neural Networks (CNNs) have already shown their great latent capacity on HRS land-cover classification in recent years. Especially, the former can achieve better performance while the latter is able to generate lightweight architecture. Unfortunately, they both have shortcomings. On the one hand, because manual CNNs are almost proposed for natural image processing, it becomes very redundant and inefficient to process HRS images. On the other hand, nascent Neural Architecture Search (NAS) techniques for dense prediction tasks are mainly based on encoder-decoder architecture, and just focus on the automatic design of the encoder, which makes it still difficult to recover the refined mapping when confronting complicated HRS scenes.To overcome their defects and tackle the HRS land-cover classification problems better, we propose AutoLC which combines the advantages of two methods. First, we devise a hierarchical search space and gain the lightweight encoder underlying gradient-based search strategy. Second, we meticulously design a lightweight but top-performing decoder that is adaptive to the searched encoder of itself. Finally, experimental results on the LoveDA land-cover dataset demonstrate that our AutoLC method outperforms the state-of-art manual and automatic methods with much less computational consumption.

With the development of remote sensing, a large amount of high-spatial-resolution (HSR) images is available, which makes refined land-cover mapping possible. However, the details of ground objects in HSR images are complex, especially in edges, therefore brings new challenges in land-cover classification. Existing deep learning method views it as a semantic segmentation task based on the fully convolutional networks (FCN), taking no account of complex details recognition. In this paper, we tackle this problem by proposing a point representation network (PointNet) for HSR land-cover classification Specifically, the uncertain point selection is designed for finding the most uncertain details at the end of ResNet encoder. According to these points, the coarse and fine features in the encoder are fused, followed by a multilayer perceptron (MLP). Different from convolutional sampling, the MLP focuses on the recognition of uncertain points, which modifies the network optimization during training. During the prediction, the coarse features are successively upsampled with the points refining, improving the performance on land-cover details recognition. Experimental results on a Nanjing land-cover dataset demonstrate that the PointNet outperforms the state-of-the-art methods.

In this paper we introduce WASP, a general communication layer for hybrid wireless networks where multiple networks are used to complement each other. In our system, we capitalize on an infrastructure with a ubiquitous,wide-area network to help enable the creation of a local mobile ad-hocnetwork in an efficient, scalable, evolvable, and manageable way. In particular, in an architecture inspired by software-defined networking,we decouple the control plane and data plane in the mobile devices and shift the control plane to a centralized controller. The controller, reachable via the wide-area network, manages a collection of mobile devices by informing each device how to handle traffic based on neighbor information provided by the mobile devices. With this, a mobile ad-hoc network can help reduce the data burden on the ubiquitous network, and the ubiquitous network can help reduce the burden on the mobile devices. WASP can be used in different networks with different applications such as cellular and military networks. In this paper, we based our implementation on Android and tested on a collection of Google Nexus-4 devices to measure metrics such as battery consumption. We evaluate on an extended ns-3 simulation platform which we added the ability to run unmodified Android applications on the nodes within ns-3. Our experiments show that WASP scales better than traditional ad-hoc networks with only a minimal trade off of energy. Additionally, we show that a content distribution scheme using WASP on smart phones with cellular data plans significantly offloads bandwidth from the cellular infrastructure, and in turn reduces expensive data usage and energy usage.

Our dependence on mobile communications continues to grow; however, wireless devices may encounter inadequate coverage due to a variety of shortage and outage circumstances. In particular, in disaster situations, the access to communications is critical for rescue operations while local wireless infrastructure may have suffered damage. In this paper, we demonstrate the concept of hybrid cellular mobile ad hoc network (hybrid cellular-MANET) to extend wireless coverage by showing a fully functioning microblogging system for smartphones and tablets without requiring modifications of the existing wireless infrastructure in an emulated disaster area. Participating devices connect to each other through WiFi radio to form a hybrid cellular-MANET, relaying data off the MANET through the nodes that have sufficient cellular forwarding capability. We have proposed a self-organizing, mobility-aware, multi-path routing protocol (HMANET) to control the data forwarding in the hybrid cellular-MANET. We compare its performance with the classic Hybrid Wireless Mesh Protocol (HWMP) used in the IEEE 802.11s standard in the built microblogging system. Our experimental results show that the HMANET protocol statistically outperforms the HWMP protocol in terms of adaptiveness to role change and mobility in the hybrid cellular-MANET.