An effective low-rank compression with a joint rank selection followed by a compression-friendly training.

Low-rank compression of a neural network is one of the popular compression techniques, where it has been known to have two main challenges. The first challenge is determining the optimal rank of all the layers and the second is training the neural network into a compression-friendly form. To overcome the two challenges, we propose BSR (Beam-search and Stable Rank), a low-rank compression algorithm that embodies an efficient rank-selection method and a unique compression-friendly training method. For the rank selection, BSR employs a modified beam search that can perform a joint optimization of the rank allocations over all the layers in contrast to the previously used heuristic methods. For the compression-friendly training, BSR adopts a regularization loss derived from a modified stable rank, which can control the rank while incurring almost no harm in performance. Experiment results confirm that BSR is effective and superior when compared to the existing low-rank compression methods. For CIFAR10 on ResNet56, BSR not only achieves compression but also provides a performance improvement over the baseline model's performance for the compression ratio of up to 0.82. For CIFAR100 on ResNet56 and ImageNet on AlexNet, BSR outperforms the previous SOTA method, LC, by 4.7% and by 6.7% on the average, respectively. BSR is also effective for EfficientNet-B0 and MobileNetV2 that are known for their efficient design in terms of parameters and computational cost. We also show that BSR provides a competitive performance when compared with the recent pruning compression algorithms. As with pruning, BSR can be easily combined with quantization for an additional compression.

Semantic Network Analysis Using Construction Accident Cases to Understand Workers' Unsafe Acts.

Unsafe acts by workers are a direct cause of accidents in the labor-intensive construction industry. Previous studies have reviewed past accidents and analyzed their causes to understand the nature of the human error involved. However, these studies focused their investigations on only a small number of construction accidents, even though a large number of them have been collected from various countries. Consequently, this study developed a semantic network analysis (SNA) model that uses approximately 60,000 construction accident cases to understand the nature of the human error that affects safety in the construction industry. A modified human factor analysis and classification system (HFACS) framework was used to classify major human error factors-that is, the causes of the accidents in each of the accident summaries in the accident case data-and an SNA analysis was conducted on all of the classified data to analyze correlations between the major factors that lead to unsafe acts. The results show that an overwhelming number of accidents occurred due to unintended acts such as perceptual errors (PERs) and skill-based errors (SBEs). Moreover, this study visualized the relationships between factors that affected unsafe acts based on actual construction accident case data, allowing for an intuitive understanding of the major keywords for each of the factors that lead to accidents.

Intracranial Efficacy of Selpercatinib in RET Fusion-Positive Non-Small Cell Lung Cancers on the LIBRETTO-001 Trial.

PURPOSE: We report the intracranial efficacy of selpercatinib, a highly potent and selective RET inhibitor, approved in the United States for RET fusion-positive non-small cell lung cancers (NSCLC). PATIENTS AND METHODS: In the global phase 1/2 LIBRETTO-001 trial (NCT03157128) in advanced RET-altered solid tumors, selpercatinib was dosed orally (160 mg twice every day) in 28-day cycles. Patients with baseline intracranial metastases had MRI/CT scans every 8 weeks for 1 year (12 weeks thereafter). In this pre-planned analysis of patients with RET fusion-positive NSCLC with baseline intracranial metastases, the primary endpoint was independently assessed intracranial objective response rate (ORR) per RECIST 1.1. Secondary endpoints included intracranial disease control rate, intracranial duration of response, and intracranial progression-free survival (PFS) independently reviewed. RESULTS: Eighty patients with NSCLC had brain metastases at baseline. Patients were heavily pretreated (median = 2 systemic therapies, range = 0-10); 56% of patients received ≥1 course of intracranial radiation (14% whole brain radiotherapy, 45% stereotactic radiosurgery). Among 22 patients with measurable intracranial disease at baseline, intracranial ORR was 82% [95% confidence interval (CI), 60-95], including 23% with complete responses. Among all intracranial responders (measurable and nonmeasurable, n = 38), median duration of intracranial response was not reached (95% CI, 9.3-NE) at a median duration of follow-up of 9.5 months (IQR = 5.7, 12.0). At 12 months, 55% of intracranial responses were ongoing. In all 80 patients, median intracranial PFS was 13.7 months (95% CI, 10.9-NE) at a median duration of follow-up of 11.0 months (IQR = 7.4, 16.5). No new safety signals were revealed in patients with brain metastases compared with the full NSCLC trial population. CONCLUSIONS: Selpercatinib has robust and durable intracranial efficacy in patients with RET fusion-positive NSCLC.

Phase 1 cohort expansion study of LY3023414, a dual PI3K/mTOR inhibitor, in patients with advanced mesothelioma.

BACKGROUND LY3023414 is a selective, ATP competitive inhibitor of class I PI3K isoforms, mTORC1/2 and DNA-PK. A Phase 1 dose escalation, 200 mg twice daily (BID) of LY3023414 was the determined recommended phase 2 dose (RP2D). We report the antitumor activity and safety of LY3023414 monotherapy in patients with advanced mesothelioma.METHODS Patients enrolled had advanced malignant pleural or peritoneal mesothelioma with measurable disease, ECOG PS 0-1, were refractory or ineligible to receive standard therapies. Patients received LY3023414 200 mg BID. This dose expansion cohort is intended to evaluate preliminary antitumor activity of LY3023414 by overall response rate. Safety, tolerability and pharmacokinetics were assessed. Biomarkers associated with treatment response was an exploratory endpoint. RESULTS Forty-two patients received LY3023414 for a median duration of 11.2 weeks (range: 1.1-53.0). One patient had a confirmed partial response (PR) (ORR 2.4%). Three patients had an unconfirmed PR. Seventeen patients had stable disease (SD) (DCR 43%). Most common adverse events (AEs) included fatigue (43%), nausea (43%), decreased appetite (38%), vomiting (33%), and diarrhea (29%). AEs were mostly mild or moderate. Grade ≥ 3 AEs were reported for 21% of patients with fatigue as the most frequent event (10%). Alterations of BAP1 were identified in 11/19 patients as the most common molecular aberration, followed by SETD2 and NF2 alterations. No obvious pattern of genetic changes/mutations in single genes or pathways was associated with anti-tumor activity. CONCLUSION LY3023414 monotherapy (200 mg BID) demonstrated an acceptable and manageable safety profile with limited single-agent activity in patients with advanced mesothelioma. ClinicalTrials.gov identifier: NCT01655225; Date of registration: 19 July 2012.

Immunomodulatory Activity of a Colony-stimulating Factor-1 Receptor Inhibitor in Patients with Advanced Refractory Breast or Prostate Cancer: A Phase I Study.

PURPOSE: Tumor-associated macrophages correlate with increased invasiveness, growth, and immunosuppression. Activation of the colony-stimulating factor-1 receptor (CSF-1R) results in proliferation, differentiation, and migration of monocytes/macrophages. This phase I study evaluated the immunologic and clinical activity, and safety profile of CSF-1R inhibition with the mAb LY3022855. PATIENTS AND METHODS: Patients with advanced refractory metastatic breast cancer (MBC) or metastatic castration-resistant prostate cancer (mCRPC) were treated with LY3022855 intravenously in 6-week cycles in cohorts: (A) 1.25 mg/kg every 2 weeks (Q2W); (B) 1.0 mg/kg on weeks 1, 2, 4, and 5; (C) 100 mg once weekly; (D)100 mg Q2W. mCRPC patients were enrolled in cohorts A and B; patients with MBC were enrolled in all cohorts. Efficacy was assessed by RECIST and Prostate Cancer Clinical Trials Working Group 2 criteria. RESULTS: Thirty-four patients (22 MBC; 12 mCRPC) received ≥1 dose of LY3022855. At day 8, circulating CSF-1 levels increased and proinflammatory monocytes CD14DIMCD16BRIGHT decreased. Best RECIST response was stable disease in five patients with MBC (23%; duration, 82-302 days) and three patients with mCRPC (25%; duration, 50-124 days). Two patients with MBC (cohort A) had durable stable disease >9 months and a third patient with MBC had palpable reduction in a nontarget neck mass. Immune-related gene activation in tumor biopsies posttreatment was observed. Common any grade treatment-related adverse events were fatigue, decreased appetite, nausea, asymptomatic increased lipase, and creatine phosphokinase. CONCLUSIONS: LY3022855 was well tolerated and showed evidence of immune modulation. Clinically meaningful stable disease >9 months was observed in two patients with MBC.

ON ESTIMATION OF THE OPTIMAL TREATMENT REGIME WITH THE ADDITIVE HAZARDS MODEL.

We propose a doubly robust estimation method for the optimal treatment regime based on an additive hazards model with censored survival data. Specifically, we introduce a new semiparametric additive hazard model which allows flexible baseline covariate effects in the control group and incorporates marginal treatment effect and its linear interaction with covariates. In addition, we propose a time-dependent propensity score to construct an A-learning type of estimating equations. The resulting estimator is shown to be consistent and asymptotically normal when either the baseline effect model for covariates or the propensity score is correctly specified. The asymptotic variance of the estimator is consistently estimated using a simple resampling method. Simulation studies are conducted to evaluate the finite-sample performance of the estimators and an application to AIDS clinical trial data is also given to illustrate the methodology.

Biochemical, Microbiological, and Sensory Characteristics of Stirred Yogurt Containing Red or Green Pepper (Capsicum annuum cv. Chungyang) Juice.

Hot pepper has anti-obesity effects by controlling appetite and reducing blood fat level. To reduce the pungency of capsaicin, red or green hot pepper juice was fermented with Bacillus licheniformis SK1230. Fermented hot pepper juice was then added into yogurt at different ratios. The pH of yogurt added with hot pepper juice was decreased from 4.61 to 4.48. Titratable acidity and counts of lactic acid bacteria were increased with increasing amount of pepper juice added. However, the viscosity was decreased significantly compared to the control. On chromaticity test, when more pepper juice was added, L*-value was decreased whereas a*- and b*- values were increased significantly (p<0.05). The spectrum of antimicrobial activity of yogurt was slightly changed compared to using pepper juice. Total polyphenol contents and antioxidant activity were increased with increasing amount of pepper juice added. Stirred yogurt added with fermented red pepper juice at 3% or green pepper juice at 1% showed high scores in flavor, appearance, texture, and overall acceptance in sensory test. Yogurt added with fermented pepper juice with reduced pungency showed also good palatability during storage at 4°C. Yogurt with added hot pepper juice can be play an important role in functional food relative to anti-obesity.

Subgroup detection and sample size calculation with proportional hazards regression for survival data.

In this paper, we propose a testing procedure for detecting and estimating the subgroup with an enhanced treatment effect in survival data analysis. Here, we consider a new proportional hazard model that includes a nonparametric component for the covariate effect in the control group and a subgroup-treatment-interaction effect defined by a change plane. We develop a score-type test for detecting the existence of the subgroup, which is doubly robust against misspecification of the baseline effect model or the propensity score but not both under mild assumptions for censoring. When the null hypothesis of no subgroup is rejected, the change-plane parameters that define the subgroup can be estimated on the basis of supremum of the normalized score statistic. The asymptotic distributions of the proposed test statistic under the null and local alternative hypotheses are established. On the basis of established asymptotic distributions, we further propose a sample size calculation formula for detecting a given subgroup effect and derive a numerical algorithm for implementing the sample size calculation in clinical trial designs. The performance of the proposed approach is evaluated by simulation studies. An application to an AIDS clinical trial data is also given for illustration.

What's in a side effect? The association between pulmonary vasodilator adverse drug events and clinical outcomes in patients with pulmonary arterial hypertension.

BACKGROUND: Adverse drug events (ADEs) with pulmonary vasodilator use in pulmonary arterial hypertension (PAH) are common. ADEs may contribute to worse quality of life; however, their relationship to prognosis is unknown. The objective of this study was to determine whether common ADEs after initiating subcutaneous treprostinil were associated with prognosis in PAH. METHODS: We assembled a retrospective cohort of participants from four clinical trials of treprostinil for PAH, including 908 participants who received subcutaneous treprostinil and 243 who received placebo at the time ADEs were ascertained. The occurrences of four common early ADEs (infusion reactions, headaches, jaw pain, or gastrointestinal side effects) were assessed during the eight weeks after starting the infusion. We used Cox proportional hazards to estimate associations between ADEs and mortality. RESULTS: No ADEs related to placebo were associated with mortality. In participants who received treprostinil, infusion reactions, headaches, and jaw pain were not associated with mortality. Gastrointestinal side effects occurring during the first eight weeks following treprostinil infusion were associated with a 57% increase in the hazard of mortality (95% CI: 14-118%). This relationship was unchanged after adjusting for demographic differences and differences in baseline PAH severity. CONCLUSIONS: Gastrointestinal ADEs after starting subcutaneous treprostinil were associated with an increased risk for mortality. Increased mortality was not observed with other early ADEs or with gastrointestinal symptoms in participants who were not receiving treprostinil at the time. This hypothesis-generating association suggests ADEs may identify different phenotypes in PAH.

Efficient estimation for accelerated failure time model under case-cohort and nested case-control sampling.

Case-cohort (Prentice, 1986) and nested case-control (Thomas, 1977) designs have been widely used as a cost-effective alternative to the full-cohort design. In this article, we propose an efficient likelihood-based estimation method for the accelerated failure time model under case-cohort and nested case-control designs. An EM algorithm is developed to maximize the likelihood function and a kernel smoothing technique is adopted to facilitate the estimation in the M-step of the EM algorithm. We show that the proposed estimators for the regression coefficients are consistent and asymptotically normal. The asymptotic variance of the estimators can be consistently estimated using an EM-aided numerical differentiation method. Simulation studies are conducted to evaluate the finite-sample performance of the estimators and an application to a Wilms tumor data set is also given to illustrate the methodology.