As the number of distributed services (or microservices) of cloud-native applications grows, resource management becomes a challenging task. These applications tend to be user-facing and latency-sensitive, and our goal is to continuously minimize the amount of CPU resources allocated while still satisfying the application latency SLO. Although previous efforts have proposed simple heuristics and sophisticated ML-based techniques, we believe that a practical resource manager should accurately scale CPU resources for diverse applications, with minimum human efforts and operation overheads. To this end, we ask: can we systematically break resource management down to subproblems solvable by practical policies? Based on the notion of CPU-throttle-based performance target, we decouple the mechanisms of SLO feedback and resource control, and implement a two-level framework -- Autothrottle. It combines a lightweight learned controller at the global level, and agile per-microservice controllers at the local level. We evaluate Autothrottle on three microservice applications, with both short-term and 21-day production workload traces. Empirical results show Autothrottle's superior CPU core savings up to 26.21% over the best-performing baselines across applications, while maintaining the latency SLO.

This is a brief technical report of our proposed method for Multiple-Object Tracking (MOT) Challenge in Complex Environments. In this paper, we treat the MOT task as a two-stage task including human detection and trajectory matching. Specifically, we designed an improved human detector and associated most of detection to guarantee the integrity of the motion trajectory. We also propose a location-wise matching matrix to obtain more accurate trace matching. Without any model merging, our method achieves 66.672 HOTA and 93.971 MOTA on the DanceTrack challenge dataset.

Transfer learning on edge is challenging due to on-device limited resources. Existing work addresses this issue by training a subset of parameters or adding model patches. Developed with inference in mind, Inverted Residual Blocks (IRBs) split a convolutional layer into depthwise and pointwise convolutions, leading to more stacking layers, e.g., convolution, normalization, and activation layers. Though they are efficient for inference, IRBs require that additional activation maps are stored in memory for training weights for convolution layers and scales for normalization layers. As a result, their high memory cost prohibits training IRBs on resource-limited edge devices, and making them unsuitable in the context of transfer learning. To address this issue, we present MobileTL, a memory and computationally efficient on-device transfer learning method for models built with IRBs. MobileTL trains the shifts for internal normalization layers to avoid storing activation maps for the backward pass. Also, MobileTL approximates the backward computation of the activation layer (e.g., Hard-Swish and ReLU6) as a signed function which enables storing a binary mask instead of activation maps for the backward pass. MobileTL fine-tunes a few top blocks (close to output) rather than propagating the gradient through the whole network to reduce the computation cost. Our method reduces memory usage by 46% and 53% for MobileNetV2 and V3 IRBs, respectively. For MobileNetV3, we observe a 36% reduction in floating-point operations (FLOPs) when fine-tuning 5 blocks, while only incurring a 0.6% accuracy reduction on CIFAR10. Extensive experiments on multiple datasets demonstrate that our method is Pareto-optimal (best accuracy under given hardware constraints) compared to prior work in transfer learning for edge devices.

In this paper, we study the problem of visual grounding by considering both phrase extraction and grounding (PEG). In contrast to the previous phrase-known-at-test setting, PEG requires a model to extract phrases from text and locate objects from images simultaneously, which is a more practical setting in real applications. As phrase extraction can be regarded as a $1$D text segmentation problem, we formulate PEG as a dual detection problem and propose a novel DQ-DETR model, which introduces dual queries to probe different features from image and text for object prediction and phrase mask prediction. Each pair of dual queries is designed to have shared positional parts but different content parts. Such a design effectively alleviates the difficulty of modality alignment between image and text (in contrast to a single query design) and empowers Transformer decoder to leverage phrase mask-guided attention to improve performance. To evaluate the performance of PEG, we also propose a new metric CMAP (cross-modal average precision), analogous to the AP metric in object detection. The new metric overcomes the ambiguity of Recall@1 in many-box-to-one-phrase cases in phrase grounding. As a result, our PEG pre-trained DQ-DETR establishes new state-of-the-art results on all visual grounding benchmarks with a ResNet-101 backbone. For example, it achieves $91.04\%$ and $83.51\%$ in terms of recall rate on RefCOCO testA and testB with a ResNet-101 backbone. Code will be availabl at \url{https://github.com/IDEA-Research/DQ-DETR}.

Self-supervised pre-training recently demonstrates success on large-scale multimodal data, and state-of-the-art contrastive learning methods often enforce the feature consistency from cross-modality inputs, such as video/audio or video/text pairs. Despite its convenience to formulate and leverage in practice, such cross-modality alignment (CMA) is only a weak and noisy supervision, since two modalities can be semantically misaligned even they are temporally aligned. For example, even in the commonly adopted instructional videos, a speaker can sometimes refer to something that is not visually present in the current frame; and the semantic misalignment would only be more unpredictable for the raw videos from the internet. We conjecture that might cause conflicts and biases among modalities, and may hence prohibit CMA from scaling up to training with larger and more heterogeneous data. This paper first verifies our conjecture by observing that, even in the latest VATT pre-training using only instructional videos, there exist strong gradient conflicts between different CMA losses within the same video, audio, text triplet, indicating them as the noisy source of supervision. We then propose to harmonize such gradients, via two techniques: (i) cross-modality gradient realignment: modifying different CMA loss gradients for each sample triplet, so that their gradient directions are more aligned; and (ii) gradient-based curriculum learning: leveraging the gradient conflict information on an indicator of sample noisiness, to develop a curriculum learning strategy to prioritize training on less noisy sample triplets. Applying those techniques to pre-training VATT on the HowTo100M dataset, we consistently improve its performance on different downstream tasks. Moreover, we are able to scale VATT pre-training to more complicated non-narrative Youtube8M dataset to further improve the state-of-the-arts.

Virtual reality (VR) over wireless is expected to be one of the killer applications in next-generation communication networks. Nevertheless, the huge data volume along with stringent requirements on latency and reliability under limited bandwidth resources makes untethered wireless VR delivery increasingly challenging. Such bottlenecks, therefore, motivate this work to seek the potential of using semantic communication, a new paradigm that promises to significantly ease the resource pressure, for efficient VR delivery. To this end, we propose a novel framework, namely WIreless SEmantic deliveRy for VR (WiserVR), for delivering consecutive 360{\deg} video frames to VR users. Specifically, deep learning-based multiple modules are well-devised for the transceiver in WiserVR to realize high-performance feature extraction and semantic recovery. Among them, we dedicatedly develop a concept of semantic location graph and leverage the joint-semantic-channel-coding method with knowledge sharing to not only substantially reduce communication latency, but also to guarantee adequate transmission reliability and resilience under various channel states. Moreover, implementation of WiserVR is presented, followed by corresponding initial simulations for performance evaluation compared with benchmarks. Finally, we discuss several open issues and offer feasible solutions to unlock the full potential of WiserVR.

持续学习（CL）依次学习像人类这样的新任务，其目标是实现更好的稳定性（S，记住过去的任务）和可塑性（P，适应新任务）。由于过去的培训数据不可用，因此探索培训示例中S和P的影响差异很有价值，这可能会改善对更好的SP的学习模式。受影响函数的启发（如果），我们首先研究了示例通过添加扰动来示例体重和计算影响推导的影响。为了避免在神经网络中Hessian逆的存储和计算负担，我们提出了一种简单而有效的METASP算法，以模拟IF计算中的两个关键步骤，并获得S-和P-Aware示例的影响。此外，我们建议通过解决双目标优化问题来融合两种示例影响，并获得对SP Pareto最优性的融合影响。融合影响可用于控制模型的更新并优化排练的存储。经验结果表明，我们的算法在任务和类别基准CL数据集上都显着优于最先进的方法。

步态冻结（FOG）是帕金森氏病的最常见症状之一，这是中枢神经系统的神经退行性疾病，影响了世界各地数百万的人。为了满足提高雾的治疗质量的紧迫需求，设计雾计算机辅助检测和量化工具的需求越来越重要。作为一种用于收集运动模式的非侵入性技术，从压力敏感步态垫中获得的脚步压力序列为评估诊所和家庭环境中的雾气提供了绝佳的机会。在这项研究中，提出了雾检测为一项顺序建模任务，并提出了一种新颖的深度学习结构，即对对抗性时空网络（ASTN），提出了跨多个级别的雾模式。引入了一种新型的对抗训练方案，并具有多级主题鉴别器，以获得独立的雾代表示，这有助于降低由于高主体间方差而导致的过度拟合风险。结果，对于看不见的受试者，可以实现强大的雾检测。拟议的计划还阐明了从其他场景中改善主题级临床研究，因为它可以与许多现有的深层建筑集成在一起。据我们所知，这是基于脚步压力的雾检测的最早研究之一，利用ASTN的方法是追求独立于主题的表示形式的第一个深神经网络架构。从21名受试者收集的393次试验的实验结果表明，AUC 0.85的雾检测提出的ASTN表现令人鼓舞。

实时音乐伴奏的生成在音乐行业（例如音乐教育和现场表演）中具有广泛的应用。但是，自动实时音乐伴奏的产生仍在研究中，并且经常在逻辑延迟和暴露偏见之间取决于权衡。在本文中，我们提出了Song Driver，这是一种无逻辑延迟或暴露偏见的实时音乐伴奏系统。具体而言，Songdriver将一个伴奏的生成任务分为两个阶段：1）安排阶段，其中变压器模型首先安排了和弦，以实时进行输入旋律，并在下一阶段加速了和弦，而不是播放它们。 2）预测阶段，其中CRF模型基于先前缓存的和弦生成了即将到来的旋律的可播放的多轨伴奏。通过这种两相策略，歌手直接生成即将到来的旋律的伴奏，从而达到了零逻辑延迟。此外，在预测时间步的和弦时，歌手是指第一阶段的缓存和弦，而不是其先前的预测，这避免了暴露偏见问题。由于输入长度通常在实时条件下受到限制，因此另一个潜在的问题是长期顺序信息的丢失。为了弥补这一缺点，我们在当前时间步骤作为全球信息之前从长期音乐作品中提取了四个音乐功能。在实验中，我们在一些开源数据集上训练歌手，以及由中国风格的现代流行音乐得分构建的原始\```````'''aisong数据集。结果表明，歌手在客观和主观指标上均优于现有的SOTA（最先进）模型，同时大大降低了物理潜伏期。

上下文自适应熵模型的应用显着提高了速率 - 渗透率（R-D）的性能，在该表现中，超级培训和自回归模型被共同利用来有效捕获潜在表示的空间冗余。但是，潜在表示仍然包含一些空间相关性。此外，这些基于上下文自适应熵模型的方法在解码过程中无法通过并行计算设备，例如FPGA或GPU。为了减轻这些局限性，我们提出了一个学识渊博的多分辨率图像压缩框架，该框架利用了最近开发的八度卷积，以将潜在表示形式分配到高分辨率（HR）和低分辨率（LR）部分，类似于小波变换，这进一步改善了R-D性能。为了加快解码的速度，我们的方案不使用上下文自适应熵模型。取而代之的是，我们利用一个额外的超层，包括超级编码器和超级解码器，以进一步删除潜在表示的空间冗余。此外，将跨分辨率参数估计（CRPE）引入提出的框架中，以增强信息流并进一步改善速率延伸性能。提出了对总损耗函数提出的其他信息损失，以调整LR部分对最终位流的贡献。实验结果表明，与最先进的学术图像压缩方法相比，我们的方法分别将解码时间减少了约73.35％和93.44％，R-D性能仍然优于H.266/VVC（4：4：：4：： 2：0）以及对PSNR和MS-SSIM指标的一些基于学习的方法。