亚洲男人的天堂2018av,欧美草比,久久久久久免费视频精选,国色天香在线看免费,久久久久亚洲av成人片仓井空

In this work, we study how the performance of a given direction changes with its sampling ratio in Multilingual Neural Machine Translation (MNMT). By training over 200 multilingual models with various model sizes, data sizes, and language directions, we find it interesting that the performance of certain translation direction does not always improve with the increase of its weight in the multi-task optimization objective. Accordingly, scalarization method leads to a multitask trade-off front that deviates from the traditional Pareto front when there exists data imbalance in the training corpus, which poses a great challenge to improve the overall performance of all directions. Based on our observations, we propose the Double Power Law to predict the unique performance trade-off front in MNMT, which is robust across various languages, data adequacy, and the number of tasks. Finally, we formulate the sample ratio selection problem in MNMT as an optimization problem based on the Double Power Law. In our experiments, it achieves better performance than temperature searching and gradient manipulation methods with only 1/5 to 1/2 of the total training budget. We release the code at //github.com/pkunlp-icler/ParetoMNMT for reproduction.

In order to perform multimodal fusion of heterogeneous signals, we need to understand their interactions: how each modality individually provides information useful for a task and how this information changes in the presence of other modalities. In this paper, we perform a comparative study of how humans annotate two categorizations of multimodal interactions: (1) partial labels, where different annotators annotate the label given the first, second, and both modalities, and (2) counterfactual labels, where the same annotator annotates the label given the first modality before asking them to explicitly reason about how their answer changes when given the second. We further propose an alternative taxonomy based on (3) information decomposition, where annotators annotate the degrees of redundancy: the extent to which modalities individually and together give the same predictions, uniqueness: the extent to which one modality enables a prediction that the other does not, and synergy: the extent to which both modalities enable one to make a prediction that one would not otherwise make using individual modalities. Through experiments and annotations, we highlight several opportunities and limitations of each approach and propose a method to automatically convert annotations of partial and counterfactual labels to information decomposition, yielding an accurate and efficient method for quantifying multimodal interactions.

Floods can cause horrific harm to life and property. However, they can be mitigated or even avoided by the effective use of hydraulic structures such as dams, gates, and pumps. By pre-releasing water via these structures in advance of extreme weather events, water levels are sufficiently lowered to prevent floods. In this work, we propose FIDLAR, a Forecast Informed Deep Learning Architecture, achieving flood management in watersheds with hydraulic structures in an optimal manner by balancing out flood mitigation and unnecessary wastage of water via pre-releases. We perform experiments with FIDLAR using data from the South Florida Water Management District, which manages a coastal area that is highly prone to frequent storms and floods. Results show that FIDLAR performs better than the current state-of-the-art with several orders of magnitude speedup and with provably better pre-release schedules. The dramatic speedups make it possible for FIDLAR to be used for real-time flood management. The main contribution of this paper is the effective use of tools for model explainability, allowing us to understand the contribution of the various environmental factors towards its decisions.

In this note, we give very simple constructions of unique neighbor expander graphs starting from spectral or combinatorial expander graphs of mild expansion. These constructions and their analysis are simple variants of the constructions of LDPC error-correcting codes from expanders, given by Sipser-Spielman\cite{SS96} (and Tanner\cite{Tanner81}), and their analysis. We also show how to obtain expanders with many unique neighbors using similar ideas. There were many exciting results on this topic recently, starting with Asherov-Dinur\cite{AD23} and Hsieh-McKenzie-Mohanty-Paredes\cite{HMMP23}, who gave a similar construction of unique neighbor expander graphs, but using more sophisticated ingredients (such as almost-Ramanujan graphs) and a more involved analysis. Subsequent beautiful works of Cohen-Roth-TaShma\cite{CRT23} and Golowich\cite{Golowich23} gave even stronger objects (lossless expanders), but also using sophisticated ingredients. The main contribution of this work is that we get much more elementary constructions of unique neighbor expanders and with a simpler analysis.

Lately, there have been intensive studies on strengths and limitations of nonuniform families of promise decision problems solvable by various types of polynomial-size finite automata families, where "polynomial-size" refers to the polynomially-bounded state complexity of a finite automata family. In this line of study, we further expand the scope of these studies to families of partial counting and gap functions, defined in terms of nonuniform families of polynomial-size nondeterministic finite automata, and their relevant families of promise decision problems. Counting functions have an ability of counting the number of accepting computation paths produced by nondeterministic finite automata. With no unproven hardness assumption, we show numerous separations and collapses of complexity classes of those partial counting and gap function families and their induced promise decision problem families. We also investigate their relationships to pushdown automata families of polynomial stack-state complexity.

In this work, we introduce Semantic Pyramid AutoEncoder (SPAE) for enabling frozen LLMs to perform both understanding and generation tasks involving non-linguistic modalities such as images or videos. SPAE converts between raw pixels and interpretable lexical tokens (or words) extracted from the LLM's vocabulary. The resulting tokens capture both the semantic meaning and the fine-grained details needed for visual reconstruction, effectively translating the visual content into a language comprehensible to the LLM, and empowering it to perform a wide array of multimodal tasks. Our approach is validated through in-context learning experiments with frozen PaLM 2 and GPT 3.5 on a diverse set of image understanding and generation tasks. Our method marks the first successful attempt to enable a frozen LLM to generate image content while surpassing state-of-the-art performance in image understanding tasks, under the same setting, by over 25%.

In this article, we propose using network-based sampling strategies to estimate the number of unsheltered people experiencing homelessness within a given administrative service unit, known as a Continuum of Care. We demonstrate the effectiveness of network sampling methods to solve this problem. Here, we focus on Respondent Driven Sampling (RDS), which has been shown to provide unbiased or low-biased estimates of totals and proportions for hard-to-reach populations in contexts where a sampling frame (e.g., housing addresses) is not available. To make the RDS estimator work for estimating the total number of people living unsheltered, we introduce a new method that leverages administrative data from the HUD-mandated Homeless Management Information System (HMIS). The HMIS provides high-quality counts and demographics for people experiencing homelessness who sleep in emergency shelters. We then demonstrate this method using network data collected in Nashville, TN, combined with simulation methods to illustrate the efficacy of this approach and introduce a method for performing a power analysis to find the optimal sample size in this setting. We conclude with the RDS unsheltered PIT count conducted by King County Regional Homelessness Authority in 2022 (data publicly available on the HUD website) and perform a comparative analysis between the 2022 RDS estimate of unsheltered people experiencing homelessness and an ARIMA forecast of the visual unsheltered PIT count. Finally, we discuss how this method works for estimating the unsheltered population of people experiencing homelessness and future areas of research.

In the evolving landscape of cybersecurity, the utilization of cyber deception has gained prominence as a proactive defense strategy against sophisticated attacks. This paper presents a comprehensive survey that investigates the crucial network requirements essential for the successful implementation of effective cyber deception techniques. With a focus on diverse network architectures and topologies, we delve into the intricate relationship between network characteristics and the deployment of deception mechanisms. This survey provides an in-depth analysis of prevailing cyber deception frameworks, highlighting their strengths and limitations in meeting the requirements for optimal efficacy. By synthesizing insights from both theoretical and practical perspectives, we contribute to a comprehensive understanding of the network prerequisites crucial for enabling robust and adaptable cyber deception strategies.

In this paper, we comprehensively investigate the potential misuse of modern Large Language Models (LLMs) for generating credible-sounding misinformation and its subsequent impact on information-intensive applications, particularly Open-Domain Question Answering (ODQA) systems. We establish a threat model and simulate potential misuse scenarios, both unintentional and intentional, to assess the extent to which LLMs can be utilized to produce misinformation. Our study reveals that LLMs can act as effective misinformation generators, leading to a significant degradation in the performance of ODQA systems. To mitigate the harm caused by LLM-generated misinformation, we explore three defense strategies: prompting, misinformation detection, and majority voting. While initial results show promising trends for these defensive strategies, much more work needs to be done to address the challenge of misinformation pollution. Our work highlights the need for further research and interdisciplinary collaboration to address LLM-generated misinformation and to promote responsible use of LLMs.

Graph Neural Networks (GNNs) have recently become increasingly popular due to their ability to learn complex systems of relations or interactions arising in a broad spectrum of problems ranging from biology and particle physics to social networks and recommendation systems. Despite the plethora of different models for deep learning on graphs, few approaches have been proposed thus far for dealing with graphs that present some sort of dynamic nature (e.g. evolving features or connectivity over time). In this paper, we present Temporal Graph Networks (TGNs), a generic, efficient framework for deep learning on dynamic graphs represented as sequences of timed events. Thanks to a novel combination of memory modules and graph-based operators, TGNs are able to significantly outperform previous approaches being at the same time more computationally efficient. We furthermore show that several previous models for learning on dynamic graphs can be cast as specific instances of our framework. We perform a detailed ablation study of different components of our framework and devise the best configuration that achieves state-of-the-art performance on several transductive and inductive prediction tasks for dynamic graphs.