Search for: All records

The eXtreme Multi-label Classification (XMC) problem seeks to find relevant labels from an exceptionally large label space. Most of the existing XMC learners focus on the extraction of semantic features from input query text. However, conventional XMC studies usually neglect the side information of instances and labels, which can be of use in many real-world applications such as recommendation systems and e-commerce product search. We propose Predicted Instance Neighborhood Aggregation (PINA), a data enhancement method for the general XMC problem that leverages beneficial side information. Unlike most existing XMC frameworks that treat labels and input instances as featureless indicators and independent entries, PINA extracts information from the label metadata and the correlations among training instances. Extensive experimental results demonstrate the consistent gain of PINA on various XMC tasks compared to the state-of-the-art methods: PINA offers a gain in accuracy compared to standard XR-Transformers on five public benchmark datasets. Moreover, PINA achieves a ∼ 5% gain in accuracy on the largest dataset LF-AmazonTitles-1.3M. Our implementation is publicly available https://github.com/amzn/pecos/ tree/mainline/examples/pina. 

We present a probabilistic framework for studying adversarial attacks on discrete data. Based on this framework, we derive a perturbation-based method, Greedy Attack, and a scalable learning-based method, Gumbel Attack, that illustrate various tradeoffs in the design of attacks. We demonstrate the effectiveness of these methods using both quantitative metrics and human evaluation on various stateof-the-art models for text classification, including a word-based CNN, a character-based CNN and an LSTM. As an example of our results, we show that the accuracy of character-based convolutional networks drops to the level of random selection by modifying only five characters through Greedy Attack. 

New results are presented on a high-statistics measurement of Collins and Sivers asymmetries of charged hadrons produced in deep inelastic scattering of muons on a transversely polarized ${}^6\mathrm{LiD}$target. The data were taken in 2022 with the COMPASS spectrometer using the 160 GeV muon beam at CERN, statistically balancing the existing data on transversely polarized proton targets. The first results from about two-thirds of the new data have total uncertainties smaller by up to a factor of three compared to the previous deuteron measurements. Using all the COMPASS proton and deuteron results, both the transversity and the Sivers distribution functions of the $u$and $d$quark, as well as the tensor charge in the measured $x$range are extracted. In particular, the accuracy of the $d$quark results is significantly improved. Published by the American Physical Society2024 

The COMPASS Collaboration performed measurements of the Drell-Yan process in 2015 and 2018 using a $190\mathrm{GeV}/c$ ${\pi }^{-}$beam impinging on a transversely polarized ammonia target. Combining the data of both years, we present final results on the amplitudes of five azimuthal modulations, which correspond to transverse-spin-dependent azimuthal asymmetries (TSAs) in the dimuon production cross section. Three of them probe the nucleon leading-twist Sivers, transversity, and pretzelosity transverse-momentum dependent (TMD) parton distribution functions (PDFs). The other two are induced by subleading effects. These TSAs provide unique new inputs for the study of the nucleon TMD PDFs and their universality properties. In particular, the Sivers TSA observed in this measurement is consistent with the fundamental QCD prediction of a sign change of naive time-reversal-odd TMD PDFs when comparing the Drell-Yan process with deep inelastic scattering. Also, within the context of model predictions, the observed transversity TSA is consistent with the expectation of a sign change for the Boer-Mulders function. Published by the American Physical Society2024