Search

Object-centric learning (OCL) aspires general and compositional understanding of scenes by representing a scene as a collection of object-centric representations. OCL has also been extended to multi-view image and video datasets to apply various data-driven inductive biases by utilizing geometric or temporal information in the multi-image data. Single-view images carry less information about how to disentangle a given scene than videos or multi-view images do. Hence, owing to the difficulty of applying inductive biases, OCL for single-view images remains challenging, resulting in inconsistent learning of object-centric representation. To this end, we introduce a novel OCL framework for single-view images, SLot Attention via SHepherding (SLASH), which consists of two simple-yet-effective modules on top of Slot Attention. The new modules, Attention Refining Kernel (ARK) and Intermediate Point Predictor and Encoder (IPPE), respectively, prevent slots from being distracted by the background noise and indicate locations for slots to focus on to facilitate learning of object-centric representation. We also propose a weak semi-supervision approach for OCL, whilst our proposed framework can be used without any assistant annotation during the inference. Experiments show that our proposed method enables consistent learning of object-centric representation and achieves strong performance across four datasets. Code is available at \url{https://github.com/object-understanding/SLASH}.

This study tests how conservatives and liberals respond differently to a trade-off between collateral penalty and collateral benefit with four different policies: two welfare-related policies and two welfare-unrelated policies.

Dense prediction tasks are a fundamental class of problems in computer vision. As supervised methods suffer from high pixel-wise labeling cost, a few-shot learning solution that can learn any dense task from a few labeled images is desired. Yet, current few-shot learning methods target a restricted set of tasks such as semantic segmentation, presumably due to challenges in designing a general and unified model that is able to flexibly and efficiently adapt to arbitrary tasks of unseen semantics. We propose Visual Token Matching (VTM), a universal few-shot learner for arbitrary dense prediction tasks. It employs non-parametric matching on patch-level embedded tokens of images and labels that encapsulates all tasks. Also, VTM flexibly adapts to any task with a tiny amount of task-specific parameters that modulate the matching algorithm. We implement VTM as a powerful hierarchical encoder-decoder architecture involving ViT backbones where token matching is performed at multiple feature hierarchies. We experiment VTM on a challenging variant of Taskonomy dataset and observe that it robustly few-shot learns various unseen dense prediction tasks. Surprisingly, it is competitive with fully supervised baselines using only 10 labeled examples of novel tasks (0.004% of full supervision) and sometimes outperforms using 0.1% of full supervision. Codes are available at https://github.com/GitGyun/visual_token_matching.

Deep Reinforcement Learning (DRL) has exhibited efficacy in resolving the Local Path Planning (LPP) problem. However, such application in the real world is immensely limited due to the deficient efficiency and generalization capability of DRL. To alleviate these two issues, a solution named Color is proposed, which consists of an Actor-Sharer-Learner (ASL) training framework and a mobile robot-oriented simulator Sparrow. Specifically, the ASL framework, intending to improve the efficiency of the DRL algorithm, employs a Vectorized Data Collection (VDC) mode to expedite data acquisition, decouples the data collection from model optimization by multithreading, and partially connects the two procedures by harnessing a Time Feedback Mechanism (TFM) to evade data underuse or overuse. Meanwhile, the Sparrow simulator utilizes a 2D grid-based world, simplified kinematics, and conversion-free data flow to achieve a lightweight design. The lightness facilitates vectorized diversity, allowing diversified simulation setups across extensive copies of the vectorized environments, resulting in a notable enhancement in the generalization capability of the DRL algorithm being trained. Comprehensive experiments, comprising 57 benchmark video games, 32 simulated and 36 real-world LPP scenarios, have been conducted to corroborate the superiority of our method in terms of efficiency and generalization. The code and the video of the experiments can be accessed on our website., Comment: 12 pages, 11 figures, 2 tables

The unfortunate fact of consumption is that enjoyment decreases with repeated consumption, a phenomenon known as hedonic decline. This work investigates how the structure of repeated consumption influences such hedonic decline. Repeated consumption may occur across a series of consumption units, such as individual songs within a playlist or a series of posts on a social media feed. This work examines how two specific consumption structures, increasing and decreasing, can influence hedonic decline for such multi-unit consumption episodes. Increasing (vs. decreasing) structures refer to circumstances where consumption quantity grows (vs. declines) with repetition (e.g., the duration of songs: 3-minutes, 4-minutes, 5-minutes (or the reverse); or the number of images in successive social media posts: 1 photo followed by 3 photos (or the reverse)). Five experiments and one archival study show that, holding the total consumption quantity constant, decreasing structures attenuate hedonic decline relative to increasing structures. This structure effect is driven by changes in beliefs about subjective quantity consumed and is mitigated by providing an external reference about consumption quantity. Further, this structure effect has significant impacts on consumer choices. These findings speak to how consumers and firms can structure consumption episodes to minimize hedonic decline and, thus, maximize utility.

Trajectory optimization (TO) aims to find a sequence of valid states while minimizing costs. However, its fine validation process is often costly due to computationally expensive collision searches, otherwise coarse searches lower the safety of the system losing a precise solution. To resolve the issues, we introduce a new collision-distance estimator, GraphDistNet, that can precisely encode the structural information between two geometries by leveraging edge feature-based convolutional operations, and also efficiently predict a batch of collision distances and gradients through 25,000 random environments with a maximum of 20 unforeseen objects. Further, we show the adoption of attention mechanism enables our method to be easily generalized in unforeseen complex geometries toward TO. Our evaluation show GraphDistNet outperforms state-of-the-art baseline methods in both simulated and real world tasks., Comment: 8 pages, 7 figures, RA-L 2022 published version

The purpose of this paper is to examine the opportunities and barriers of Integrated Human-Machine Intelligence (IHMI) in civil engineering. Integrating artificial intelligence's high efficiency and repeatability with humans' adaptability in various contexts can advance timely and reliable decision-making during civil engineering projects and emergencies. Successful cases in other domains, such as biomedical science, healthcare, and transportation, showed the potential of IHMI in data-driven, knowledge-based decision-making in numerous civil engineering applications. However, whether the industry and academia are ready to embrace the era of IHMI and maximize its benefit to the industry is still questionable due to several knowledge gaps. This paper thus calls for future studies in exploring the value, method, and challenges of applying IHMI in civil engineering. Our systematic review of the literature and motivating cases has identified four knowledge gaps in achieving effective IHMI in civil engineering. First, it is unknown what types of tasks in the civil engineering domain can be assisted by AI and to what extent. Second, the interface between human and AI in civil engineering-related tasks need more precise and formal definition. Third, the barriers that impede collecting detailed behavioral data from humans and contextual environments deserve systematic classification and prototyping. Lastly, it is unknown what expected and unexpected impacts will IHMI have on the AEC industry and entrepreneurship. Analyzing these knowledge gaps led to a list of identified research questions. This paper will lay the foundation for identifying relevant studies to form a research roadmap to address the four knowledge gaps identified.

Generic Event Boundary Detection (GEBD) is a newly suggested video understanding task that aims to find one level deeper semantic boundaries of events. Bridging the gap between natural human perception and video understanding, it has various potential applications, including interpretable and semantically valid video parsing. Still at an early development stage, existing GEBD solvers are simple extensions of relevant video understanding tasks, disregarding GEBD's distinctive characteristics. In this paper, we propose a novel framework for unsupervised/supervised GEBD, by using the Temporal Self-similarity Matrix (TSM) as the video representation. The new Recursive TSM Parsing (RTP) algorithm exploits local diagonal patterns in TSM to detect boundaries, and it is combined with the Boundary Contrastive (BoCo) loss to train our encoder to generate more informative TSMs. Our framework can be applied to both unsupervised and supervised settings, with both achieving state-of-the-art performance by a huge margin in GEBD benchmark. Especially, our unsupervised method outperforms the previous state-of-the-art "supervised" model, implying its exceptional efficacy.

We present a generalization of Transformers to any-order permutation invariant data (sets, graphs, and hypergraphs). We begin by observing that Transformers generalize DeepSets, or first-order (set-input) permutation invariant MLPs. Then, based on recently characterized higher-order invariant MLPs, we extend the concept of self-attention to higher orders and propose higher-order Transformers for order-$k$ data ($k=2$ for graphs and $k>2$ for hypergraphs). Unfortunately, higher-order Transformers turn out to have prohibitive complexity $\mathcal{O}(n^{2k})$ to the number of input nodes $n$. To address this problem, we present sparse higher-order Transformers that have quadratic complexity to the number of input hyperedges, and further adopt the kernel attention approach to reduce the complexity to linear. In particular, we show that the sparse second-order Transformers with kernel attention are theoretically more expressive than message passing operations while having an asymptotically identical complexity. Our models achieve significant performance improvement over invariant MLPs and message-passing graph neural networks in large-scale graph regression and set-to-(hyper)graph prediction tasks. Our implementation is available at https://github.com/jw9730/hot., 21 pages, 5 figures

The correlation between the structural phase transition (SPT) and oxygen vacancy in SrRuO3 (SRO) thin films was investigated by in situ X-ray diffraction (XRD) and ambient pressure X-ray photoelectron spectroscopy (AP-XPS). In situ XRD shows that the SPT occurs from a monoclinic SRO phase to a tetragonal SRO phase near ∼200 °C, regardless of the pressure environment. On the other hand, significant core level shifts in both the Ru and Sr photoemission spectra are found under ultrahigh vacuum, but not under the oxygen pressure environment. The directions and behavior of the core level shift of Ru and Sr are attributed to the formation of oxygen vacancy across the SPT temperature of SRO. The analysis of in situ XRD and AP-XPS results provides an evidence for the formation of metastable surface oxide possibly due to the migration of internal oxygen atoms across the SPT temperature, indicating the close relationship between oxygen vacancy and SPT in SRO thin films.

We characterize Pareto optimality via "near" weighted utilitarian welfare maximization. One characterization sequentially maximizes utilitarian welfare functions using a finite sequence of nonnegative and eventually positive welfare weights. The other maximizes a utilitarian welfare function with a certain class of positive hyperreal weights. The social welfare ordering represented by these "near" weighted utilitarian welfare criteria is characterized by the standard axioms for weighted utilitarianism under a suitable weakening of the continuity axiom.

Beamforming is the core principle used for communication in mmWave bands to overcome the excessive pathloss experienced at these bands. Usage of narrow beams results in a large number of beams at transmitter and receiver covering the given range of azimuth and elevation angles. Narrow beams require frequent beam alignment to ensure maximum beam gain of the link, requiring periodic beam search. Beam search complexity is proportional to the number of transmit-receive beam pairs. The larger beam-pair set will inherently delay the full search and hence the delay in finding the best possible beam pair. In this paper, we analyze the beam selection algorithm at user mobile equipment (UE) with device orientation change. We model a propagation channel with device orientation changes using the 3GPP channel model and understand the effect of it. Further, we propose a beam tracking and selection algorithm using orientation sensors to optimize the best beam selection procedure by reducing the beam search space. We show a possible practical implementation of the proposed algorithm on the actual mmWave device. Performance evaluation of the proposed algorithm is done both on the simulator and in the lab setup in various conditions. We illustrate the gains in downlink by throughput and in uplink by transmit power. For example, with this proposed algorithm, at 50Âo/sec orientation change rate, downlink throughput gain is 33% and the uplink power gain is 6 dB. Similarly, lab results show a throughput gain of 20% in the downlink with typical human usage scenarios.

This paper introduces the concept of a non-uniform constellation (NUC) in contrast to conventional uniform quadrature-amplitude modulation (QAM) constellations. Such constellations provide additional shaping gain, which allows reception at lower signal-to-noise ratios. ATSC3.0 will be the first major broadcasting standard, which completely uses NUCs due to their outstanding properties. We will consider different kinds of NUCs and describe their performance: 2-D NUCs provide more shaping gain at the cost of higher demapping complexity, while 1-D NUCs allow low-complexity demapping at slightly lower shaping gains. These NUCs are well suited for very large constellations sizes, such as 1k and 4k QAM.

In this paper, we introduce the overview and detailed structure of low-density parity-check (LDPC) codes, which are adopted for the physical layer standard of Advanced Television Systems Committee (ATSC) 3.0. As one of the features to mitigate channel errors efficiently, LDPC codes are used as a main part of forward error correction scheme within the bit interleaved coded modulation chain. This paper presents two different structures of LDPC codes in the ATSC 3.0 standard: 1) irregular repeat accumulate structure and 2) multi-edge type structure. After the detailed encoding methods for the two structures are introduced, the performance results including comparison studies with other broadcasting standards such as DVB-T2 and DVB-S2 are shown.

The physical layer [layer-1 (L1)] signaling (L1 signaling) of Advanced Television Systems Committee (ATSC) 3.0 is studied. In the new digital broadcasting system, ATSC 3.0, L1 signaling is designed to be more efficient but still robust compared with preceding standards such as DVB-T2. We introduce the error protection scheme of ATSC 3.0 in detail. Especially advanced methods newly employed in ATSC 3.0 are investigated with regard to its purpose and benefit in comparison to conventional DVB-T2.

We develop a network-flow approach for characterizing interim-allocation rules that can be implemented by ex post allocations. Our method can be used to characterize feasible interim allocations in general multi-unit auctions where agents face capacity constraints, both ceilings and floors. Applications include a variety of settings of practical interest, ranging from individual and group-specific capacity constraints, set-aside sale, partnership dissolution, and government license reallocation.

Peter DeMarzo, Ilan Kremer and Andrzej Skrzypacz (2005, henceforth DKS) analyzed auctions in which bidders compete in securities. They show that a steeper security leads to a higher expected revenue for the seller, and also use this to establish the revenue ranking between standard auctions. In this comment, we obtain the opposite results to DKS's by assuming that a higher return requires a higher investment cost. Given this latter assumption, steeper securities are more vulnerable to adverse selection, and may yield lower expected revenue, than flatter ones.

We describe the design of a digital noticeboard to support communication within a remote Aboriginal community whose aspiration is to live in "both worlds", nurturing and extending their Aboriginal culture and actively participating in Western society and economy. Three bi-cultural aspects have emerged and are presented here: the need for a bi-lingual noticeboard to span both oral and written language traditions, the tension between perfunctory information exchange and social, embodied protocols of telling in person and the different ways in which time is represented in both cultures. The design approach, developed iteratively through consultation, demonstration and testing led to an "unsurprising interface", aimed at maximizing use and appropriation across cultures by unifying visual, text and spoken contents in both passive and interactive displays in a modeless manner.