Extracting informative arguments of events from news articles is a challenging problem in information extraction, which requires a global contextual understanding of each document. While recent work on document-level extraction has gone beyond single-sentence and increased the cross-sentence inference capability of end-to-end models, they are still restricted by certain input sequence length constraints and usually ignore the global context between events. To tackle this issue, we introduce a new global neural generation-based framework for document-level event argument extraction by constructing a document memory store to record the contextual event information and leveraging it to implicitly and explicitly help with decoding of arguments for later events. Empirical results show that our framework outperforms prior methods substantially and it is more robust to adversarially annotated examples with our constrained decoding design. (Our code and resources are available at https://github.com/xinyadu/memory_docie for research purpose.)

Trajectory prediction is essential for autonomous vehicles (AVs) to plan correct and safe driving behaviors. While many prior works aim to achieve higher prediction accuracy, few study the adversarial robustness of their methods. To bridge this gap, we propose to study the adversarial robustness of data-driven trajectory prediction systems. We devise an optimization-based adversarial attack framework that leverages a carefully-designed differentiable dynamic model to generate realistic adversarial trajectories. Empirically, we benchmark the adversarial robustness of state-of-the-art prediction models and show that our attack increases the prediction error for both general metrics and planning-aware metrics by more than 50% and 37%. We also show that our attack can lead an AV to drive off road or collide into other vehicles in simulation. Finally, we demonstrate how to mitigate the adversarial attacks using an adversarial training scheme.

Traditional statistical methods include the time-domain Time-series analysis is used to extracting meaningful methods, such as the family of autoregressive (AR) models statistics and characteristics of temporal sequences and their many variants, including ARMA (AR moving of data [1], and is among the most ubiquitous mathematical average), ARIMA (AR integrated moving average), methods. Indeed, time-series are universal for SARIMA (seasonal ARIMA), etc. [1]. Such models use a signal processing methods and in pattern recognition applications, diversity of optimization techniques to estimate parameters dominating characterization of econometrics of a linear model with its history dependence. Traditional and finance along with almost any scientific and engineering frequency-domain methods use the properties of application. Time-series methods can be broadly short-time Fourier transforms [9] and/or wavelet transforms divided into time-domain and frequency-domain methods, [10] in order to characterize the signal in a joint the former of which uses a variety of statistical techniques time-frequency representation. More recently, there have to characterize a sequence, and the latter of which been efforts to model time-series data as from a dynamical uses spectral (e.g.

A large body of work shows that machine learning (ML) models can leak sensitive or confidential information about their training data. Recently, leakage due to distribution inference (or property inference) attacks is gaining attention. In this attack, the goal of an adversary is to infer distributional information about the training data. So far, research on distribution inference has focused on demonstrating successful attacks, with little attention given to identifying the potential causes of the leakage and to proposing mitigations. To bridge this gap, as our main contribution, we theoretically and empirically analyze the sources of information leakage that allows an adversary to perpetrate distribution inference attacks. We identify three sources of leakage: (1) memorizing specific information about the $\mathbb{E}[Y|X]$ (expected label given the feature values) of interest to the adversary, (2) wrong inductive bias of the model, and (3) finiteness of the training data. Next, based on our analysis, we propose principled mitigation techniques against distribution inference attacks. Specifically, we demonstrate that causal learning techniques are more resilient to a particular type of distribution inference risk termed distributional membership inference than associative learning methods. And lastly, we present a formalization of distribution inference that allows for reasoning about more general adversaries than was previously possible.

Federated learning, where algorithms are trained across multiple decentralized devices without sharing local data, is increasingly popular in distributed machine learning practice. Typically, a graph structure $G$ exists behind local devices for communication. In this work, we consider parameter estimation in federated learning with data distribution and communication heterogeneity, as well as limited computational capacity of local devices. We encode the distribution heterogeneity by parametrizing distributions on local devices with a set of distinct $p$-dimensional vectors. We then propose to jointly estimate parameters of all devices under the $M$-estimation framework with the fused Lasso regularization, encouraging an equal estimate of parameters on connected devices in $G$. We provide a general result for our estimator depending on $G$, which can be further calibrated to obtain convergence rates for various specific problem setups. Surprisingly, our estimator attains the optimal rate under certain graph fidelity condition on $G$, as if we could aggregate all samples sharing the same distribution. If the graph fidelity condition is not met, we propose an edge selection procedure via multiple testing to ensure the optimality. To ease the burden of local computation, a decentralized stochastic version of ADMM is provided, with convergence rate $O(T^{-1}\log T)$ where $T$ denotes the number of iterations. We highlight that, our algorithm transmits only parameters along edges of $G$ at each iteration, without requiring a central machine, which preserves privacy. We further extend it to the case where devices are randomly inaccessible during the training process, with a similar algorithmic convergence guarantee. The computational and statistical efficiency of our method is evidenced by simulation experiments and the 2020 US presidential election data set.

Two Chinese astronauts went on a spacewalk Saturday from a new space station that is due to be completed later this year. Cai Xuzhe and Chen Dong's installed pumps, a handle to open the hatch door from outside in an emergency, and a foot-stop to fix an astronaut's feet to a robotic arm, state media said. China is building its own space station after being excluded by the U.S. from the International Space Station because its military runs the country's space program. American officials see a host of strategic challenges from China's space ambitions, in an echo of the U.S.-Soviet rivalry that prompted the race to the moon in the 1960s. The latest spacewalk was the second during a six-month mission that will oversee the completion of the space station.

Spanning from Silicon Valley to Upper Manhattan and everywhere in between, it seems this country is overflowing with tech-optimists. As you know -- a techno-optimist is someone who is generally optimistic about the current state of technology and its potential future. These people believe that technological developments will do more good for humanity than harm -- and that our technological future is very bright. These Techno-Optimists may believe that technology has the power to solve major crises, like global climate change, or believe that machine learning and AI will enable us to reach incredible new heights in humanity. They also tend to envision a technologically rich future, with science fiction-inspired gadgets and capabilities in the hands of average people.

"We do not want schizophrenia researchers knowing a lot about software engineering," said Amy Winecoff, data scientist and Princeton's Centre for IT Policy. Research asserts that a basic understanding of machine learning and other software engineering principles might be a desirable trait for medical practitioners, but these skills should not come at the expense of expertise in domain knowledge. Many new startups and enterprises sell their products boasting about incorporating AI/ML techniques in the development. Though this is an issue in the developer and business market, the bigger worry is misapplied AI/ML algorithms in the field of science and healthcare as it causes real world consequences. Sayash Kapoor and Arvind Narayanan of Princeton University published a research paper--Leakage and the Reproducibility Crisis in ML-based Science, pointing out the problem of "data leakage" in various researches using pools of data to train and test their development's performance.

Representative image San Francisco: The US space agency has collaborated with Google to help local governments improve their monitoring and prediction of air quality. NASA and Google will develop advanced machine learning-based algorithms that link space data with Google Earth Engine data streams to generate high-resolution air quality maps in near real-time. "We're thrilled about our partnership with NASA to make daily air quality more actionable at a local level," said Rebecca Moore, director at Google Earth, Earth Engine and Outreach at Google. The results will create city-scale, near real-time estimation and forecasting of harmful pollutants, such as nitrogen dioxide and fine particulate matter. Google has incorporated two new NASA data sets into the Earth Engine Catalogue that are automatically updated daily.

Artificial intelligence expert Stela Solar joins Olivia Shen and Dr Will Stoltz to discuss the strategic implications of AI technology.