No abstract

No abstract

Prompt-based methods have been successfully applied in sentence-level few-shot learning tasks, mostly owing to the sophisticated design of templates and label words. However, when applied to token-level labeling tasks such as NER, it would be time-consuming to enumerate the template queries over all potential entity spans. In this work, we propose a more elegant method to reformulate NER tasks as LM problems without any templates. Specifically, we discard the template construction process while maintaining the word prediction paradigm of pre-training models to predict a class-related pivot word (or label word) at the entity position. Meanwhile, we also explore principled ways to automatically search for appropriate label words that the pre-trained models can easily adapt to. While avoiding the complicated template-based process, the proposed LM objective also reduces the gap between different objectives used in pre-training and fine-tuning, thus it can better benefit the few-shot performance. Experimental results demonstrate the effectiveness of the proposed method over bert-tagger and template-based method under few-shot settings. Moreover, the decoding speed of the proposed method is up to 1930.12 times faster than the template-based method.

Primary organic aerosol (POA) emitted from light duty gasoline vehicles (LDGVs) exhibits a semivolatile behavior in which heating the aerosol and/or diluting the aerosol leads to partial evaporation of the POA. A single volatility distribution can explain the median evaporation behavior of POA emitted from LDGVs but this approach is unable to capture the full range of measured POA volatility during thermodenuder (TD) experiments conducted at atmospherically relevant concentrations (2-5 μg m(-3)). Reanalysis of published TD data combined with analysis of new measurements suggest that POA emitted from gasoline vehicles is composed of two types of POA that have distinctly different volatility distributions: one low-volatility distribution and one medium-volatility distribution. These correspond to fuel combustion-derived POA and motor oil POA, respectively. Models that simultaneously incorporate both of these distributions are able to reproduce experimental results much better (R(2) = 0.94) than models that use a single average or median distribution (R(2) = 0.52). These results indicate that some fraction of POA emitted from LDGVs is essentially nonvolatile under typical atmospheric dilution levels. Roughly 50% of the vehicles tested in the current study had POA emissions dominated by fuel combustion products (essentially nonvolatile). Further testing is required to determine appropriate fleet-average emissions rates of the two POA types from LDGVs.

The problem of state estimation in the setting of partially-observed discrete event systems subject to cyber attacks is considered. An operator observes a plant through a natural projection that hides the occurrence of certain events. The objective of the operator is that of estimating the current state of the system. The observation is corrupted by an attacker which can tamper with the readings of a set of sensors thus inserting some fake events or erasing some observations. The aim of the attacker is that of altering the state estimation of the operator. An automaton, called attack structure, is defined to describe the set of all possible attacks. In more details, an unbounded attack structure is obtained by concurrent composition of two state observers, the attacker observer and the operator observer. The attack structure shows, for each possible corrupted observation, the joint state estimation, i.e., the set of states consistent with the uncorrupted observation and the set of states consistent with the corrupted observation. Such a structure can be used to establish if an attack function is harmful w.r.t. a misleading relation. Our approach is also extended to the case in which the attacker may insert at most n events between two consecutive observations.INDEX TERMS Discrete event systems, state estimation, cyber attacks.

No abstract

Anthropogenic land subsidence is an example of changes to the natural environment due to human activities and is one of the key factors in causing land degradation at a range of scales. Previous studies assessing land subsidence in the Valley of Mexico either focused on regional scale or short (noncontinuous) temporal scale. In this study, long‐term land subsidence (~15 years) is mapped in Mexico City (Mexico) using two interferometric synthetic aperture radar (InSAR) methods, namely, GEOS (Geoscience and Earth Observing Systems Group)‐Advance Time‐series Analysis and GEOS‐Small Baseline Subset. An inverse distance weighted‐based integration module and maximum likelihood regression‐based M estimator are introduced to further enhance these two methods. The land subsidence was continuously mapped using ENVISAT (2004–2007), ALOS‐1 (2007–2011), COSMO‐SkyMed (2011–2014), ALOS‐2 (2014–2018), and SENTINEL‐1 (2015–2017) data sets. A comparison between InSAR time‐series and GPS measurement shows that the subsidence rates are consistent over 2004–2018. The subsidence map over 15 years was generated finding a maximum subsidence over 4.5 m. By comparing our InSAR results with a land use map, we find that the subsidence centre in Mexico City is mostly located in the residential regions with the consumption of groundwater contributing considerably to the local subsidence rate. A total volume of 1.20 × 108 m3 of the land in Ciudad Nezahualcoyotl subsided/degraded. A continuing subsidence process limits the potential land use causing serious land degradation. Our results may be used to assist disaster reduction plans.

No abstract