现代深度学习在各个领域取得了巨大的成功。但是，它需要标记大量数据，这是昂贵且劳动密集型的。积极学习（AL）确定要标记的最有用的样本，对于最大化培训过程的效率变得越来越重要。现有的AL方法主要仅使用单个最终固定模型来获取要标记的样品。这种策略可能还不够好，因为没有考虑为给定培训数据的模型的结构不确定性来获取样品。在这项研究中，我们提出了一种基于常规随机梯度下降（SGD）优化产生的时间自我汇总的新颖获取标准。通过捕获通过SGD迭代获得的中间网络权重来获得这些自我复杂模型。我们的收购功能依赖于学生和教师模型之间的一致性度量。为学生模型提供了固定数量的时间自我安装模型，并且教师模型是通过平均学生模型来构建的。使用拟议的获取标准，我们提出了AL算法，即基于学生教师的AL（ST-Conal）。在CIFAR-10，CIFAR-100，CALTECH-256和TINY IMAGENET数据集上进行的图像分类任务进行的实验表明，所提出的ST-Conal实现的性能要比现有的获取方法要好得多。此外，广泛的实验显示了我们方法的鲁棒性和有效性。

人们对从长尾班级分布中学习的具有挑战性的视觉感知任务越来越兴趣。训练数据集中的极端类失衡使模型偏向于识别多数级数据而不是少数级数据。最近，已经提出了两个分支网络的双分支网络（DBN）框架。传统的分支和重新平衡分支用于提高长尾视觉识别的准确性。重新平衡分支使用反向采样器来生成类平衡的训练样本，以减轻由于类不平衡而减轻偏见。尽管该策略在处理偏见方面非常成功，但使用反向采样器进行培训可以降低表示形式的学习绩效。为了减轻这个问题，常规方法使用了精心设计的累积学习策略，在整个培训阶段，重新平衡分支的影响逐渐增加。在这项研究中，我们旨在开发一种简单而有效的方法，以不需要优化的累积学习而在不累积学习的情况下提高DBN的性能。我们设计了一种称为双边混合增强的简单数据增强方法，该方法将统一采样器中的一个样品与反向采样器中的另一个样品结合在一起，以产生训练样本。此外，我们介绍了阶级条件的温度缩放，从而减轻对拟议的DBN结构的多数级别的偏见。我们对广泛使用的长尾视觉识别数据集进行的实验表明，双边混合增加在改善DBN的表示性能方面非常有效，并且所提出的方法可以实现某些类别的先进绩效。

Conditional normalizing flows can generate diverse image samples for solving inverse problems. Most normalizing flows for inverse problems in imaging employ the conditional affine coupling layer that can generate diverse images quickly. However, unintended severe artifacts are occasionally observed in the output of them. In this work, we address this critical issue by investigating the origins of these artifacts and proposing the conditions to avoid them. First of all, we empirically and theoretically reveal that these problems are caused by ``exploding variance'' in the conditional affine coupling layer for certain out-of-distribution (OOD) conditional inputs. Then, we further validated that the probability of causing erroneous artifacts in pixels is highly correlated with a Mahalanobis distance-based OOD score for inverse problems in imaging. Lastly, based on our investigations, we propose a remark to avoid exploding variance and then based on it, we suggest a simple remedy that substitutes the affine coupling layers with the modified rational quadratic spline coupling layers in normalizing flows, to encourage the robustness of generated image samples. Our experimental results demonstrated that our suggested methods effectively suppressed critical artifacts occurring in normalizing flows for super-resolution space generation and low-light image enhancement without compromising performance.

We discuss a platform that has both software and hardware components, and whose purpose is to support research into characterizing and mitigating the sim-to-real gap in robotics and vehicle autonomy engineering. The software is operating-system independent and has three main components: a simulation engine called Chrono, which supports high-fidelity vehicle and sensor simulation; an autonomy stack for algorithm design and testing; and a development environment that supports visualization and hardware-in-the-loop experimentation. The accompanying hardware platform is a 1/6th scale vehicle augmented with reconfigurable mountings for computing, sensing, and tracking. Since this vehicle platform has a digital twin within the simulation environment, one can test the same autonomy perception, state estimation, or controls algorithms, as well as the processors they run on, in both simulation and reality. A demonstration is provided to show the utilization of this platform for autonomy research. Future work will concentrate on augmenting ART/ATK with support for a full-sized Chevy Bolt EUV, which will be made available to this group in the immediate future.

Image super-resolution is a common task on mobile and IoT devices, where one often needs to upscale and enhance low-resolution images and video frames. While numerous solutions have been proposed for this problem in the past, they are usually not compatible with low-power mobile NPUs having many computational and memory constraints. In this Mobile AI challenge, we address this problem and propose the participants to design an efficient quantized image super-resolution solution that can demonstrate a real-time performance on mobile NPUs. The participants were provided with the DIV2K dataset and trained INT8 models to do a high-quality 3X image upscaling. The runtime of all models was evaluated on the Synaptics VS680 Smart Home board with a dedicated edge NPU capable of accelerating quantized neural networks. All proposed solutions are fully compatible with the above NPU, demonstrating an up to 60 FPS rate when reconstructing Full HD resolution images. A detailed description of all models developed in the challenge is provided in this paper.

Model-Based Reinforcement Learning (RL) is widely believed to have the potential to improve sample efficiency by allowing an agent to synthesize large amounts of imagined experience. Experience Replay (ER) can be considered a simple kind of model, which has proved extremely effective at improving the stability and efficiency of deep RL. In principle, a learned parametric model could improve on ER by generalizing from real experience to augment the dataset with additional plausible experience. However, owing to the many design choices involved in empirically successful algorithms, it can be very hard to establish where the benefits are actually coming from. Here, we provide theoretical and empirical insight into when, and how, we can expect data generated by a learned model to be useful. First, we provide a general theorem motivating how learning a model as an intermediate step can narrow down the set of possible value functions more than learning a value function directly from data using the Bellman equation. Second, we provide an illustrative example showing empirically how a similar effect occurs in a more concrete setting with neural network function approximation. Finally, we provide extensive experiments showing the benefit of model-based learning for online RL in environments with combinatorial complexity, but factored structure that allows a learned model to generalize. In these experiments, we take care to control for other factors in order to isolate, insofar as possible, the benefit of using experience generated by a learned model relative to ER alone.

Recently, numerous studies have investigated cooperative traffic systems using the communication among vehicle-to-everything (V2X). Unfortunately, when multiple autonomous vehicles are deployed while exposed to communication failure, there might be a conflict of ideal conditions between various autonomous vehicles leading to adversarial situation on the roads. In South Korea, virtual and real-world urban autonomous multi-vehicle races were held in March and November of 2021, respectively. During the competition, multiple vehicles were involved simultaneously, which required maneuvers such as overtaking low-speed vehicles, negotiating intersections, and obeying traffic laws. In this study, we introduce a fully autonomous driving software stack to deploy a competitive driving model, which enabled us to win the urban autonomous multi-vehicle races. We evaluate module-based systems such as navigation, perception, and planning in real and virtual environments. Additionally, an analysis of traffic is performed after collecting multiple vehicle position data over communication to gain additional insight into a multi-agent autonomous driving scenario. Finally, we propose a method for analyzing traffic in order to compare the spatial distribution of multiple autonomous vehicles. We study the similarity distribution between each team's driving log data to determine the impact of competitive autonomous driving on the traffic environment.

We present a unified and compact representation for object rendering, 3D reconstruction, and grasp pose prediction that can be inferred from a single image within a few seconds. We achieve this by leveraging recent advances in the Neural Radiance Field (NeRF) literature that learn category-level priors and fine-tune on novel objects with minimal data and time. Our insight is that we can learn a compact shape representation and extract meaningful additional information from it, such as grasping poses. We believe this to be the first work to retrieve grasping poses directly from a NeRF-based representation using a single viewpoint (RGB-only), rather than going through a secondary network and/or representation. When compared to prior art, our method is two to three orders of magnitude smaller while achieving comparable performance at view reconstruction and grasping. Accompanying our method, we also propose a new dataset of rendered shoes for training a sim-2-real NeRF method with grasping poses for different widths of grippers.

Pretrained Language Models (LMs) memorize a vast amount of knowledge during initial pretraining, including information that may violate the privacy of personal lives and identities. Previous work addressing privacy issues for language models has mostly focused on data preprocessing and differential privacy methods, both requiring re-training the underlying LM. We propose knowledge unlearning as an alternative method to reduce privacy risks for LMs post hoc. We show that simply performing gradient ascent on target token sequences is effective at forgetting them with little to no degradation of general language modeling performances for larger LMs; it sometimes even substantially improves the underlying LM with just a few iterations. We also find that sequential unlearning is better than trying to unlearn all the data at once and that unlearning is highly dependent on which kind of data (domain) is forgotten. By showing comparisons with a previous data preprocessing method and a decoding method known to mitigate privacy risks for LMs, we show that unlearning can give a stronger empirical privacy guarantee in scenarios where the data vulnerable to extraction attacks are known a priori while being much more efficient and robust. We release the code and dataset needed to replicate our results at https://github.com/joeljang/knowledge-unlearning.

为了更快地培训深度学习模型，近年来对多个GPU的分布培训是非常受欢迎的计划。但是，沟通带宽仍然是训练表现的主要瓶颈。为了提高整体培训表现，最近的工作提出了梯度稀疏方法，可大大降低通信流量。他们中的大多数都需要梯度分类来选择有意义的梯度，例如TOP-K梯度稀疏（Top-K SGD）。但是，Top-K SGD有一个限制，可以提高速度的总体训练性能，因为梯度分类对GPU的效率显着效率低下。在本文中，我们进行的实验表明了Top-K SGD的效率低下，并提供了低性能的见解。根据我们的经验分析的观察结果，我们计划产生高性能梯度稀疏方法作为未来的工作。