Source author record

Winston Wu

Winston Wu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Catalog footprint

What is connected

2works
3topics
4close collaborators

Actions

Connect this record

Log in to claim

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

2 published item(s)

preprint2022arXiv

Trends in Silicates in the $β$ Pictoris Disk

While beta Pic is known to host silicates in ring-like structures, whether the properties of these silicate dust vary with stellocentric distance remains an open question. We re-analyze the beta Pictoris debris disk spectrum from the Spitzer Infrared Spectrograph (IRS) and a new IRTF/SpeX spectrum to investigate trends in Fe/Mg ratio, shape, and crystallinity in grains as a function of wavelength, a proxy for stellocentric distance. By analyzing a re-calibrated and re-extracted spectrum, we identify a new 18 micron forsterite emission feature and recover a 23 micron forsterite emission feature with a substantially larger line-to-continuum ratio than previously reported. We find that these prominent spectral features are primarily produced by small submicron-sized grains, which are continuously generated and replenished from planetesimal collisions in the disk and can elucidate their parent bodies' composition. We discover three trends about these small grains: as stellocentric distance increases, (1) small silicate grains become more crystalline (less amorphous), (2) they become more irregular in shape, and (3) for crystalline silicate grains, the Fe/Mg ratio decreases. Applying these trends to beta Pic's planetary architecture, we find that the dust population exterior to the orbits of beta Pic b and c differs substantially in crystallinity and shape. We also find a tentative 3-5 micron dust excess due to spatially unresolved hot dust emission close to the star. From our findings, we infer that the surfaces of large planetesimals are more Fe-rich and collisionally-processed closer to the star but more Fe-poor and primordial farther from the star.

preprint2020arXiv

Evaluating Neural Machine Comprehension Model Robustness to Noisy Inputs and Adversarial Attacks

We evaluate machine comprehension models' robustness to noise and adversarial attacks by performing novel perturbations at the character, word, and sentence level. We experiment with different amounts of perturbations to examine model confidence and misclassification rate, and contrast model performance in adversarial training with different embedding types on two benchmark datasets. We demonstrate improving model performance with ensembling. Finally, we analyze factors that effect model behavior under adversarial training and develop a model to predict model errors during adversarial attacks.