mixup: Beyond empirical risk minimization
Large deep neural networks are powerful, but exhibit undesirable behaviors such as
memorization and sensitivity to adversarial examples. In this work, we propose mixup, a …
memorization and sensitivity to adversarial examples. In this work, we propose mixup, a …
[HTML][HTML] Detection methods and clinical applications of circulating tumor cells in breast cancer
H Zhang, X Lin, Y Huang, M Wang, C Cen… - Frontiers in …, 2021 - frontiersin.org
Circulating Tumor Cells (CTCs) are cancer cells that split away from the primary tumor and
appear in the circulatory system as singular units or clusters, which was first reported by Dr …
appear in the circulatory system as singular units or clusters, which was first reported by Dr …
[HTML][HTML] LRPPRC: a multifunctional protein involved in energy metabolism and human disease
J Cui, L Wang, X Ren, Y Zhang, H Zhang - Frontiers in Physiology, 2019 - frontiersin.org
The pentatricopeptide repeat (PPR) family plays a major role in RNA stability, regulation,
processing, splicing, translation, and editing. Leucine-rich PPR-motif-containing protein …
processing, splicing, translation, and editing. Leucine-rich PPR-motif-containing protein …
[HTML][HTML] Hypoxia in breast cancer—scientific translation to therapeutic and diagnostic clinical applications
Y Zhang, H Zhang, M Wang, T Schmid, Z Xin… - Frontiers in …, 2021 - frontiersin.org
Breast cancer has been the leading cause of female cancer deaths for decades. Intratumoral
hypoxia, mainly caused by structural and functional abnormalities in microvasculature, is …
hypoxia, mainly caused by structural and functional abnormalities in microvasculature, is …
Active and Selective Conversion of CO2 to CO on Ultrathin Au Nanowires
In this communication, we show that ultrathin Au nanowires (NWs) with dominant edge sites
on their surface are active and selective for electrochemical reduction of CO2 to CO. We first …
on their surface are active and selective for electrochemical reduction of CO2 to CO. We first …
MiR-7, inhibited indirectly by lincRNA HOTAIR, directly inhibits SETDB1 and reverses the EMT of breast cancer stem cells by downregulating the STAT3 pathway
H Zhang, K Cai, J Wang, X Wang, K Cheng, F Shi… - Stem …, 2014 - academic.oup.com
Epithelial–mesenchymal transition (EMT) contributes to tumor invasion and metastasis in
many cancers and correlates highly with the acquisition of cancer stem cell (CSC) …
many cancers and correlates highly with the acquisition of cancer stem cell (CSC) …
Fixup initialization: Residual learning without normalization
Normalization layers are a staple in state-of-the-art deep neural network architectures. They
are widely believed to stabilize training, enable higher learning rate, accelerate …
are widely believed to stabilize training, enable higher learning rate, accelerate …
Riemannian SVRG: Fast stochastic optimization on Riemannian manifolds
We study optimization of finite sums of\emph {geodesically} smooth functions on
Riemannian manifolds. Although variance reduction techniques for optimizing finite-sums …
Riemannian manifolds. Although variance reduction techniques for optimizing finite-sums …
[PDF][PDF] DNA sensing in mismatch repair-deficient tumor cells is essential for anti-tumor immunity
Increased neoantigens in hypermutated cancers with DNA mismatch repair deficiency
(dMMR) are proposed as the major contributor to the high objective response rate in anti-PD …
(dMMR) are proposed as the major contributor to the high objective response rate in anti-PD …
First-order methods for geodesically convex optimization
Geodesic convexity generalizes the notion of (vector space) convexity to nonlinear metric
spaces. But unlike convex optimization, geodesically convex (g-convex) optimization is …
spaces. But unlike convex optimization, geodesically convex (g-convex) optimization is …