Short-video recommendation systems typically optimize for user satisfaction. However, allocating exposure to creators at critical growth stages incentivizes long-term content supply despite compromising immediate user engagement. Existing efforts concerning creator interests aim at either improving creator exposure fairness, matching creators with suitable audiences, or leveraging creator behavior to enhance user welfare. Directly maximizing the joint value of user satisfaction and creator incentive at recommendation time, however, remains largely unaddressed. This presents two challenges. First, the two objectives are heterogeneous in nature, making it non-trivial to formulate this joint optimization as a tractable problem. Second, optimizing creator incentive requires globally coordinated decisions across requests, making real-time serving infeasible. To address these challenges, we formulate the joint maximization as a constrained optimization problem that unifies the two heterogeneous objectives. We further derive an efficient online algorithm based on the primal-dual method, which decouples global incentive constraints into real-time decisions with theoretical guarantees. Experiments on a large-scale short-video platform demonstrate consistent improvements in joint user-creator value over existing baselines.

The rise of artificial intelligence (AI) and growing energy constraints have intensified the need for efficient, low-power hardware. Thexnorlogic, a cornerstone of low-precision neural network inference, is especially critical to optimize for energy efficiency. Conventional CMOS-based in-memoryxnorunits, however, face key limitations, including low integration density, readout interference, and multistep write protocols. To address these challenges, we present a MoS2transistor-enhanced antisymmetric 2T2R in-memoryxnorlogic unit, enabling both write and compute operations to be finished within a single step, achieving the theoretical minimum for in-memory logic primitives. Beyond logic-level optimization, the incorporation of MoS2transistors facilitates monolithic 3-D integration and lowers static power consumption relative to standard CMOS devices. At the system level, we extend the analog majority architecture to BNN inference, including the weight mapping strategy, and discuss the system-level advantages and quantization errors introduced by this approach.

In Few-Shot Learning (FSL), the objective is to correctly recognize new samples from novel classes with only a few available samples per class. Existing methods in FSL primarily focus on learning transferable knowledge from base classes by maximizing the information between feature representations and their corresponding labels. However, this approach may suffer from the "supervision collapse" issue, which arises due to a bias towards the base classes. In this paper, we propose a solution to address this issue by preserving the intrinsic structure of the data and enabling the learning of a generalized model for the novel classes. Following the InfoMax principle, our approach maximizes two types of mutual information (MI): between the samples and their feature representations, and between the feature representations and their class labels. This allows us to strike a balance between discrimination (capturing class-specific information) and generalization (capturing common characteristics across different classes) in the feature representations. To achieve this, we adopt a unified framework that perturbs the feature embedding space using two low-bias estimators. The first estimator maximizes the MI between a pair of intra-class samples, while the second estimator maximizes the MI between a sample and its augmented views. This framework effectively combines knowledge distillation between class-wise pairs and enlarges the diversity in feature representations. By conducting extensive experiments on popular FSL benchmarks, our proposed approach achieves comparable performances with state-of-the-art competitors. For example, we achieved an accuracy of 69.53% on the miniImageNet dataset and 77.06% on the CIFAR-FS dataset for the 5-way 1-shot task.

In this paper, we propose a new Momentum Contrastive Bidirectional Encoding network with S elf-D istillation (MoCoBE-SD) to alleviate the data sparsity and noise issues in sequential recommendation by providing rich informative supervisions from both sequence-level and item-level perspectives. In particular, a Momentum Contrastive Bidirectional Encoding (MoCoBE) network is first proposed by constructing momentum updated encoder based on an online bidirectional self-attention encoder, where a momentum contrastive learning task and a masked item prediction task are simultaneously optimized. Building upon MoCoBE, a well-elaborated Self-Distillation (SD) scheme is incorporated to further suppress the noise influence. Specifically, a well-trained sequence encoder by MoCoBE is adopted as the teacher encoder to provide refined supervision for the masked item prediction, which constitutes our MoCoBE-SD framework. Extensive experiments on three public datasets show that MoCoBE-SD outperforms the existing state-of-the-art methods consistently.

In this paper, we propose a new multi-scale similarity aggregation method (MSA) for dynamic metric learning (DyML), which adopts a pretraining-finetuning scheme and efficiently learns the similarity relationship for each semantic level. In particular, building upon the framework of self-supervised pretraining, the output embedding layer is divided into three learners to learn the similarity relations in each level individually. Then for training these learners, the hierarchical prior information is fully considered. Specifically, in light of the class hierarchy that each class in a coarse level corresponds to a set of subclasses in a finer level, multi-proxy learning is employed to facilitate the single-level similarity learning of each learner. On the other hand, following the hierarchical consistency property, a cross-level similarity constraint is further presented to encourage the estimated similarities of the three learners to be hierarchically consistent. Extensive experiments on three DyML datasets show that MSA significantly outperforms the existing state-of-the-art methods and allows for a better generalization for different semantic scales.

In this paper, we present a feature-augmentation approach called Masked Feature Generation Network (MFGN) for Few-Shot Learning (FSL), a challenging task that attempts to recognize the novel classes with a few visual instances for each class. Most of the feature-augmentation approaches tackle FSL tasks via modeling the intra-class distributions. We extend this idea further to explicitly capture the intra-class variations in a one-to-many manner. Specifically, MFGN consists of an encoder-decoder architecture, with an encoder that performs as a feature extractor and extracts the feature embeddings of the available visual instances (the unavailable instances are seen to be masked), along with a decoder that performs as a feature generator and reconstructs the feature embeddings of the unavailable visual instances from both the available feature embeddings and the masked tokens. Equipped with this generative architecture, MFGN produces nontrivial visual features for the novel classes with limited visual instances. In extensive experiments on four FSL benchmarks, MFGN performs competitively and outperforms the state-of-the-art competitors on most of the few-shot classification tasks.

Deep Metric Learning, a task that learns a feature embedding space where semantically similar samples are located closer than dissimilar samples, is a cornerstone of many computer vision applications. Most of the existing proxy-based approaches usually exploit the global context via learning a single proxy for each training class, which struggles in capturing the complex non-uniform data distribution with different patterns. In this work, we present an easy-to-implement framework to effectively capture the local neighbor relationships via learning multiple proxies for each class that collectively approximate the intra-class distribution. In the context of large intra-class visual diversity, we revisit the entropy learning under the multi-proxy learning framework and provide a training routine that both minimizes the entropy of intra-class probability distribution and maximizes the entropy of inter-class probability distribution. In this way, our model is able to better capture the intra-class variations and smooth the inter-class differences and thus facilitates to extract more semantic feature representations for the downstream tasks. Extensive experimental results demonstrate that the proposed approach achieves competitive performances. Codes and an appendix are provided.

Deep metric learning has attracted much attention in recent years, due to seamlessly combining the distance metric learning and deep neural network. Many endeavors are devoted to design different pair-based angular loss functions, which decouple the magnitude and direction information for embedding vectors and ensure the training and testing measure consistency. However, these traditional angular losses cannot guarantee that all the sample embeddings are on the surface of the same hypersphere during the training stage, which would result in unstable gradient in batch optimization and may influence the quick convergence of the embedding learning. In this paper, we first investigate the effect of the embedding norm for deep metric learning with angular distance, and then propose a spherical embedding constraint (SEC) to regularize the distribution of the norms. SEC adaptively adjusts the embeddings to fall on the same hypersphere and performs more balanced direction update. Extensive experiments on deep metric learning, face recognition, and contrastive self-supervised learning show that the SEC-based angular space learning strategy significantly improves the performance of the state-of-the-art.

Mobile edge computing (MEC) has become a dominant technology in the upcoming era of the 5th generation mobile networks. By offloading tasks from mobile devices to edge clouds provided by cellular base stations, both energy consumption and end-to-end delay of mobile tasks can be reduced. In this paper, we aim to optimize the latency performance of TDMA-based MEC systems by joint allocation of computation and communication resource. Our goal is to minimize the maximal delay of all devices in the system. We first simplify the optimization problem and convert it into a convex one. Then we derive the closed-form expression for the optimal resource allocation strategy and investigate the relationship between uplink and downlink resource allocation. A subgradient algorithm is also developed to solve the joint resource allocation problem. Finally, numerical simulation results are shown to verify that our proposal can achieve a better performance compared with the traditional schemes.

This paper presents a general optimization model gleaned ideas from plant root growth behaviors in the soil. The purpose of the study is to investigate a novel biologically inspired methodology for complex system modelling and computation, particularly for constrained multi-objective optimization. A novel method called “multi-objective root growth algorithm” (MORGA) for constrained multi-objective optimization is proposed based on the root growth model. A self-adaptive strategy is adopted to tie this model closer to plant root growth behaviors in nature, as well as improve the robustness of MORGA. Simulation experiments of MORGA on a set of benchmark test functions are compared with other nature inspired techniques for multi-objective optimization which includes nondominated sorting genetic algorithm II (NSGA II) and multi-objective particle swarm optimization (MOPSO). The numerical results demonstrate MORGA approach is a powerful search and optimization technique for constrained multi-objective optimization.