Filters








234,043 Hits in 1.8 sec

Adversarial Example Decomposition [article]

Horace He, Aaron Lou, Qingxuan Jiang, Isay Katsman, Serge Belongie, Ser-Nam Lim
2019 arXiv   pre-print
Research has shown that widely used deep neural networks are vulnerable to carefully crafted adversarial perturbations. Moreover, these adversarial perturbations often transfer across models. We hypothesize that adversarial weakness is composed of three sources of bias: architecture, dataset, and random initialization. We show that one can decompose adversarial examples into an architecture-dependent component, data-dependent component, and noise-dependent component and that these components
more » ... ave intuitively. For example, noise-dependent components transfer poorly to all other models, while architecture-dependent components transfer better to retrained models with the same architecture. In addition, we demonstrate that these components can be recombined to improve transferability without sacrificing efficacy on the original model.
arXiv:1812.01198v2 fatcat:vgiaxxu4ajcuxmoaa4sn2tklke

Edge Proposal Sets for Link Prediction [article]

Abhay Singh, Qian Huang, Sijia Linda Huang, Omkar Bhalerao, Horace He, Ser-Nam Lim, Austin R. Benson
2021 arXiv   pre-print
Graphs are a common model for complex relational data such as social networks and protein interactions, and such data can evolve over time (e.g., new friendships) and be noisy (e.g., unmeasured interactions). Link prediction aims to predict future edges or infer missing edges in the graph, and has diverse applications in recommender systems, experimental design, and complex systems. Even though link prediction algorithms strongly depend on the set of edges in the graph, existing approaches
more » ... ally do not modify the graph topology to improve performance. Here, we demonstrate how simply adding a set of edges, which we call a proposal set, to the graph as a pre-processing step can improve the performance of several link prediction algorithms. The underlying idea is that if the edges in the proposal set generally align with the structure of the graph, link prediction algorithms are further guided towards predicting the right edges; in other words, adding a proposal set of edges is a signal-boosting pre-processing step. We show how to use existing link prediction algorithms to generate effective proposal sets and evaluate this approach on various synthetic and empirical datasets. We find that proposal sets meaningfully improve the accuracy of link prediction algorithms based on both neighborhood heuristics and graph neural networks. Code is available at .
arXiv:2106.15810v1 fatcat:knaxp2o3yfagppej7w7hqdrkdu

Intermediate Level Adversarial Attack for Enhanced Transferability [article]

Qian Huang, Zeqi Gu, Isay Katsman, Horace He, Pian Pawakapan, Zhiqiu Lin, Serge Belongie, Ser-Nam Lim
2018 arXiv   pre-print
Neural networks are vulnerable to adversarial examples, malicious inputs crafted to fool trained models. Adversarial examples often exhibit black-box transfer, meaning that adversarial examples for one model can fool another model. However, adversarial examples may be overfit to exploit the particular architecture and feature representation of a source model, resulting in sub-optimal black-box transfer attacks to other target models. This leads us to introduce the Intermediate Level Attack
more » ... , which attempts to fine-tune an existing adversarial example for greater black-box transferability by increasing its perturbation on a pre-specified layer of the source model. We show that our method can effectively achieve this goal and that we can decide a nearly-optimal layer of the source model to perturb without any knowledge of the target models.
arXiv:1811.08458v1 fatcat:ad2t25f2mvd4fc32tpra2l3g5e

Better Set Representations For Relational Reasoning [article]

Qian Huang, Horace He, Abhay Singh, Yan Zhang, Ser-Nam Lim, Austin Benson
2020 arXiv   pre-print
Incorporating relational reasoning into neural networks has greatly expanded their capabilities and scope. One defining trait of relational reasoning is that it operates on a set of entities, as opposed to standard vector representations. Existing end-to-end approaches typically extract entities from inputs by directly interpreting the latent feature representations as a set. We show that these approaches do not respect set permutational invariance and thus have fundamental representational
more » ... tations. To resolve this limitation, we propose a simple and general network module called a Set Refiner Network (SRN). We first use synthetic image experiments to demonstrate how our approach effectively decomposes objects without explicit supervision. Then, we insert our module into existing relational reasoning models and show that respecting set invariance leads to substantial gains in prediction performance and robustness on several relational reasoning tasks.
arXiv:2003.04448v2 fatcat:mfrdthl2qjfrjow67a7fyrfdme

AMERICAN FOLK-LORE. He introduced HORACE

Howard Furness
unpublished
NEWELL, HORACE E. SCUDDER, Committee on Finance. i888. Recez5ts.  ...  HORACE E. SCUDDER, of Cambridge, Mass. A report was presented at the end of the year, embodying the financial statement submitted below.  ... 
fatcat:o52gbhdbqnepbmsnvlz5noimwm

The Pile: An 800GB Dataset of Diverse Text for Language Modeling [article]

Leo Gao, Stella Biderman, Sid Black, Laurence Golding, Travis Hoppe, Charles Foster, Jason Phang, Horace He, Anish Thite, Noa Nabeshima, Shawn Presser, Connor Leahy
2020 arXiv   pre-print
Horace He performed the bias and sentiment analysis. Anish Thite implemented and performed the profanity analysis and processed Hacker News. Noa Nabeshima processed GitHub.  ...  He wrote, however, an essay "On the External Use of Water," in which he seems to have partly anticipated the method of the cold-water cure.  ... 
arXiv:2101.00027v1 fatcat:74dgmcl55rdupks3kzygosjlca

Combining Label Propagation and Simple Models Out-performs Graph Neural Networks [article]

Qian Huang, Horace He, Abhay Singh, Ser-Nam Lim, Austin R. Benson
2020 arXiv   pre-print
Graph Neural Networks (GNNs) are the predominant technique for learning over graphs. However, there is relatively little understanding of why GNNs are successful in practice and whether they are necessary for good performance. Here, we show that for many standard transductive node classification benchmarks, we can exceed or match the performance of state-of-the-art GNNs by combining shallow models that ignore the graph structure with two simple post-processing steps that exploit correlation in
more » ... he label structure: (i) an "error correlation" that spreads residual errors in training data to correct errors in test data and (ii) a "prediction correlation" that smooths the predictions on the test data. We call this overall procedure Correct and Smooth (C&S), and the post-processing steps are implemented via simple modifications to standard label propagation techniques from early graph-based semi-supervised learning methods. Our approach exceeds or nearly matches the performance of state-of-the-art GNNs on a wide variety of benchmarks, with just a small fraction of the parameters and orders of magnitude faster runtime. For instance, we exceed the best known GNN performance on the OGB-Products dataset with 137 times fewer parameters and greater than 100 times less training time. The performance of our methods highlights how directly incorporating label information into the learning algorithm (as was done in traditional techniques) yields easy and substantial performance gains. We can also incorporate our techniques into big GNN models, providing modest gains. Our code for the OGB results is at https://github.com/Chillee/CorrectAndSmooth.
arXiv:2010.13993v2 fatcat:7tnvv2aa6rb3jewabamr4ca6pm

Enhancing Adversarial Example Transferability with an Intermediate Level Attack [article]

Qian Huang, Isay Katsman, Horace He, Zeqi Gu, Serge Belongie, Ser-Nam Lim
2020 arXiv   pre-print
Neural networks are vulnerable to adversarial examples, malicious inputs crafted to fool trained models. Adversarial examples often exhibit black-box transfer, meaning that adversarial examples for one model can fool another model. However, adversarial examples are typically overfit to exploit the particular architecture and feature representation of a source model, resulting in sub-optimal black-box transfer attacks to other target models. We introduce the Intermediate Level Attack (ILA),
more » ... attempts to fine-tune an existing adversarial example for greater black-box transferability by increasing its perturbation on a pre-specified layer of the source model, improving upon state-of-the-art methods. We show that we can select a layer of the source model to perturb without any knowledge of the target models while achieving high transferability. Additionally, we provide some explanatory insights regarding our method and the effect of optimizing for adversarial examples using intermediate feature maps. Our code is available at https://github.com/CUVL/Intermediate-Level-Attack.
arXiv:1907.10823v3 fatcat:2skaysbh6bgmrhclalnamq44dy

Incremental Kernel Ridge Regression for the Prediction of Soft Tissue Deformations [chapter]

Binbin Pan, James J. Xia, Peng Yuan, Jaime Gateno, Horace H. S. Ip, Qizhen He, Philip K. M. Lee, Ben Chow, Xiaobo Zhou
2012 Lecture Notes in Computer Science  
This paper proposes a nonlinear regression model to predict soft tissue deformation after maxillofacial surgery. The feature which served as input in the model is extracted with Finite Element Model (FEM). The output in the model is the facial deformation calculated from the preoperative and postoperative 3D data. After finding the relevance between feature and facial deformation by using the regression model, we establish a general relationship which can be applied to all the patients. As a
more » ... patient comes, we predict his/her facial deformation by combining the general relationship and the new patient's biomechanical properties. Thus, our model is biomechanical relevant and statistical relevant. Validation on eleven patients demonstrates the effectiveness and efficiency of our method.
doi:10.1007/978-3-642-33415-3_13 fatcat:buzjihlphbgb3bk5pqvb7bahxm

Neurod1 Modulates Opioid Antinociceptive Tolerance via Two Distinct Mechanisms

Wen Li, Songwei He, Yuye Zhou, Yuan Li, Jianbang Hao, Xingru Zhou, Feng Wang, Yang Zhang, Zhenhua Huang, Zhiyuan Li, Horace H. Loh, Ping-Yee Law (+1 others)
2014 Biological Psychiatry  
Background-The activity of neurogenic differentiation 1 (Neurod1) decreases after morphine administration, which leads to impairments of the stability of dendritic spines in primary hippocampal neurons, adult neurogenesis in mouse hippocampi, and drug-associated contextual memory. The current study examined whether Neurod1 could affect the development of opioid tolerance. Neurod1" or short hairpin RNA against Neurod1 was injected into mouse hippocampi separately or combined (more than eight
more » ... for each treatment) to modulate Neurod1 activity. The antinociceptive median effective dose values of morphine and fentanyl were determined with tail-flick assay and used to calculate development of tolerance. Contextual learning and memory were assayed using the Morris water maze. Methods-Lentivirus encoding Results-Decrease in NeuroD1 activity increased the initial antinociceptive median effective dose values of both morphine and fentanyl, which was reversed by restoring NeuroD1 activity. In contrast, decrease in NeuroD1 activity inhibited development of tolerance in a time-dependent manner, paralleling its effects on the acquisition and extinction of contextual memory. In addition, only development of tolerance, but not antinociceptive median effective dose values, was modulated by the expression of miR-190 and Neurod1 driven by Nestin promoter. Conclusions-Neurod1 regulates the developments of opioid tolerance via a time-dependent pathway through contextual learning and a short-response pathway through antinociception.
doi:10.1016/j.biopsych.2014.05.013 pmid:24993058 pmcid:PMC4503258 fatcat:jtbdsnpvujgvllyhpei57tewwm

Prediction of soft tissue deformations after CMF surgery with incremental kernel ridge regression

Binbin Pan, Guangming Zhang, James J. Xia, Peng Yuan, Horace H.S. Ip, Qizhen He, Philip K.M. Lee, Ben Chow, Xiaobo Zhou
2016 Computers in Biology and Medicine  
Facial soft tissue deformation following osteotomy is associated with the corresponding biomechanical characteristics of bone and soft tissues. However, none of the methods devised to predict soft tissue deformation after osteotomy incorporates population-based statistical data. The aim of this study is to establish a statistical model to describe the relationship between biomechanical characteristics and soft tissue deformation after osteotomy. We proposed an incremental kernel ridge
more » ... (IKRR) model to accomplish this goal. The input of the model is the biomechanical information computed by the Finite Element Method (FEM). The output is the soft tissue deformation generated from the paired pre-operative and post-operative 3D images. The model is adjusted incrementally with each new patient's biomechanical information. Therefore, the IKRR model enables us to predict potential soft tissue deformations for new patient by using both biomechanical and statistical information. The integration of these two types of data is critically important for accurate simulations of soft-tissue changes after surgery. The proposed method was evaluated by leave-one-out cross-validation using data from 11 patients. The average prediction error of our model (0.9103 mm) was lower than some state-of-the-art algorithms. This model is promising as a reliable way to prevent the risk of facial distortion after craniomaxillofacial surgery.
doi:10.1016/j.compbiomed.2016.04.020 pmid:27213920 pmcid:PMC5279917 fatcat:g3qcs6ihkjaadd2ia7yvpu4jby

Measuring Coding Challenge Competence With APPS [article]

Dan Hendrycks and Steven Basart and Saurav Kadavath and Mantas Mazeika and Akul Arora and Ethan Guo and Collin Burns and Samir Puranik and Horace He and Dawn Song and Jacob Steinhardt
2021 arXiv   pre-print
If the current player can't choose any number satisfying the conditions, he loses. Can you determine the winner if they both play optimally?  ...  In each player's turn, he has to choose an integer a and subtract it from n such that: 1 ≤ a ≤ n. If it's Mahmoud's turn, a has to be even, but if it's Ehab's turn, a has to be odd.  ... 
arXiv:2105.09938v3 fatcat:yz47dmlk2jfg7j7vruckphmv34

Incidence and predictors of Lhermitte's sign among patients receiving mediastinal radiation for lymphoma

Bassem Youssef, JoAnn Shank, Jay P. Reddy, Chelsea C. Pinnix, George Farha, Mani Akhtari, Pamela K. Allen, Michelle A. Fanale, John A. Garcia, Patricia H. Horace, Sarah Milgrom, Grace Li Smith (+6 others)
2015 Radiation Oncology  
Purpose: To prospectively examine the risk of developing Lhermitte's sign (LS) in patients with lymphoma treated with modern-era chemotherapy followed by consolidation intensity-modulated radiation therapy. Methods: We prospectively interviewed all patients with lymphoma who received irradiation to the mediastinum from July 2011 through April 2014. We extracted patient, disease, and treatment-related variables from the medical records of those patients and dosimetric variables from
more » ... nning systems and analyzed these factors to identify potential predictors of LS with Pearson chi-square tests. Results: During the study period 106 patients received mediastinal radiation for lymphoma, and 31 (29 %) developed LS. No correlations were found between LS and any of the variables examined, including total radiation dose, maximum point dose to the spinal cord, volume receiving 105 % of the dose, and volumes receiving 5 or 15 Gy. Conclusion: In this group of patients, treatment with chemotherapy followed by intensity-modulated radiation therapy led to 29 % developing LS; this symptom was independent of radiation dose and seemed to be an idiosyncratic reaction. This relatively high incidence could have resulted from prospective use of a structured interview.
doi:10.1186/s13014-015-0504-7 pmid:26407853 pmcid:PMC4582821 fatcat:ygpxkc3guvb2vnv63dofvpc72a

Opioid doses required for pain management in lung cancer patients with different cholesterol levels: negative correlation between opioid doses and cholesterol levels

Zhenhua Huang, Lining Liang, Lingyu Li, Miao Xu, Xiang Li, Hao Sun, Songwei He, Lilong Lin, Yixin Zhang, Yancheng Song, Man Yang, Yuling Luo (+4 others)
2016 Lipids in Health and Disease  
Pain management has been considered as significant contributor to broad quality-of-life improvement for cancer patients. Modulating serum cholesterol levels affects analgesia abilities of opioids, important pain killer for cancer patients, in mice system. Thus the correlation between opioids usages and cholesterol levels were investigated in human patients with lung cancer. Methods: Medical records of 282 patients were selected with following criteria, 1) signed inform consent, 2) full medical
more » ... ecords on total serum cholesterol levels and opioid administration, 3) opioid-naïve, 4) not received/ receiving cancer-related or cholesterol lowering treatment, 5) pain level at level 5-8. The patients were divided into different groups basing on their gender and cholesterol levels. Since different opioids, morphine, oxycodone, and fentanyl, were all administrated at fixed low dose initially and increased gradually only if pain was not controlled, the percentages of patients in each group who did not respond to the initial doses of opioids and required higher doses for pain management were determined and compared. Results: Patients with relative low cholesterol levels have larger percentage (11 out of 28 in female and 31 out of 71 in male) to not respond to the initial dose of opioids than those with high cholesterol levels (0 out of 258 in female and 8 out of 74 in male). Similar differences were obtained when patients with different opioids were analyzed separately. After converting the doses of different opioids to equivalent doses of oxycodone, significant correlation between opioid usages and cholesterol levels was also observed. Conclusions: Therefore, more attention should be taken to those cancer patients with low cholesterol levels because they may require higher doses of opioids as pain killer.
doi:10.1186/s12944-016-0212-9 pmid:26952011 pmcid:PMC4782347 fatcat:sbgqrwoyorafxpmrw4akeowgu4

Clinical, Virological and Immunological Features from Patients Infected with Re-Emergent Avian-Origin Human H7N9 Influenza Disease of Varying Severity in Guangdong Province

Zi Feng Yang, Chris Ka Pun Mok, Xiao Qing Liu, Xiao Bo Li, Jian Feng He, Wen Da Guan, Yong Hao Xu, Wei Qi Pan, Li Yan Chen, Yong Ping Lin, Shi Guan Wu, Si Hua Pan (+16 others)
2015 PLoS ONE  
Importantly, he had detectable HSV DNA in the throat swabs from day 23 post onset (CT 30.63) and the viral load progressively increased (CT 18.05) by day 29 of illness.  ...  H) and the cytokine levels in the plasma samples of the survived patients declined along with the improvement of their condition while dramatic increase of cytokines were observed in patient 3 before he  ... 
doi:10.1371/journal.pone.0117846 pmid:25723593 pmcid:PMC4344233 fatcat:tbnz7vaagjcajlyiefahbq4mby
« Previous Showing results 1 — 15 out of 234,043 results