Showing papers by "New York University published in 2015"

Deep learning

Abstract: Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.

Purposeful Sampling for Qualitative Data Collection and Analysis in Mixed Method Implementation Research.

Abstract: Purposeful sampling is widely used in qualitative research for the identification and selection of information-rich cases related to the phenomenon of interest. Although there are several different purposeful sampling strategies, criterion sampling appears to be used most commonly in implementation research. However, combining sampling strategies may be more appropriate to the aims of implementation research and more consistent with recent developments in quantitative methods. This paper reviews the principles and practice of purposeful sampling in implementation research, summarizes types and categories of purposeful sampling strategies and provides a set of recommendations for use of single strategy or multistage strategy designs, particularly for state implementation research.

Estimating the reproducibility of psychological science

Abstract: Reproducibility is a defining feature of science, but the extent to which it characterizes current research is unknown. We conducted replications of 100 experimental and correlational studies published in three psychology journals using high-powered designs and original materials when available. Replication effects were half the magnitude of original effects, representing a substantial decline. Ninety-seven percent of original studies had statistically significant results. Thirty-six percent of replications had statistically significant results; 47% of original effect sizes were in the 95% confidence interval of the replication effect size; 39% of effects were subjectively rated to have replicated the original result; and if no bias in original results is assumed, combining original and replication results left 68% with statistically significant effects. Correlational tests suggest that replication success was better predicted by the strength of original evidence than by characteristics of the original and replication teams.

Highly Parallel Genome-wide Expression Profiling of Individual Cells Using Nanoliter Droplets

Abstract: Cells, the basic units of biological structure and function, vary broadly in type and state. Single-cell genomics can characterize cell identity and function, but limitations of ease and scale have prevented its broad application. Here we describe Drop-seq, a strategy for quickly profiling thousands of individual cells by separating them into nanoliter-sized aqueous droplets, associating a different barcode with each cell's RNAs, and sequencing them all together. Drop-seq analyzes mRNA transcripts from thousands of individual cells simultaneously while remembering transcripts' cell of origin. We analyzed transcriptomes from 44,808 mouse retinal cells and identified 39 transcriptionally distinct cell populations, creating a molecular atlas of gene expression for known retinal cell classes and novel candidate cell subtypes. Drop-seq will accelerate biological discovery by enabling routine transcriptional profiling at single-cell resolution. VIDEO ABSTRACT.

The eleventh and twelfth data releases of the Sloan Digital Sky Survey: final data from SDSS-III

Abstract: The third generation of the Sloan Digital Sky Survey (SDSS-III) took data from 2008 to 2014 using the original SDSS wide-field imager, the original and an upgraded multi-object fiber-fed optical spectrograph, a new near-infrared high-resolution spectrograph, and a novel optical interferometer. All the data from SDSS-III are now made public. In particular, this paper describes Data Release 11 (DR11) including all data acquired through 2013 July, and Data Release 12 (DR12) adding data acquired through 2014 July (including all data included in previous data releases), marking the end of SDSS-III observing. Relative to our previous public release (DR10), DR12 adds one million new spectra of galaxies and quasars from the Baryon Oscillation Spectroscopic Survey (BOSS) over an additional 3000 sq. deg of sky, more than triples the number of H-band spectra of stars as part of the Apache Point Observatory (APO) Galactic Evolution Experiment (APOGEE), and includes repeated accurate radial velocity measurements of 5500 stars from the Multi-Object APO Radial Velocity Exoplanet Large-area Survey (MARVELS). The APOGEE outputs now include measured abundances of 15 different elements for each star. In total, SDSS-III added 2350 sq. deg of ugriz imaging; 155,520 spectra of 138,099 stars as part of the Sloan Exploration of Galactic Understanding and Evolution 2 (SEGUE-2) survey; 2,497,484 BOSS spectra of 1,372,737 galaxies, 294,512 quasars, and 247,216 stars over 9376 sq. deg; 618,080 APOGEE spectra of 156,593 stars; and 197,040 MARVELS spectra of 5,513 stars. Since its first light in 1998, SDSS has imaged over 1/3 of the Celestial sphere in five bands and obtained over five million astronomical spectra.

The Global Burden of Cancer 2013

Abstract: Importance Cancer is among the leading causes of death worldwide. Current estimates of cancer burden in individual countries and regions are necessary to inform local cancer control strategies. Objective To estimate mortality, incidence, years lived with disability (YLDs), years of life lost (YLLs), and disability-adjusted life-years (DALYs) for 28 cancers in 188 countries by sex from 1990 to 2013. Evidence Review The general methodology of the Global Burden of Disease (GBD) 2013 study was used. Cancer registries were the source for cancer incidence data as well as mortality incidence (MI) ratios. Sources for cause of death data include vital registration system data, verbal autopsy studies, and other sources. The MI ratios were used to transform incidence data to mortality estimates and cause of death estimates to incidence estimates. Cancer prevalence was estimated using MI ratios as surrogates for survival data; YLDs were calculated by multiplying prevalence estimates with disability weights, which were derived from population-based surveys; YLLs were computed by multiplying the number of estimated cancer deaths at each age with a reference life expectancy; and DALYs were calculated as the sum of YLDs and YLLs. Findings In 2013 there were 14.9 million incident cancer cases, 8.2 million deaths, and 196.3 million DALYs. Prostate cancer was the leading cause for cancer incidence (1.4 million) for men and breast cancer for women (1.8 million). Tracheal, bronchus, and lung (TBL) cancer was the leading cause for cancer death in men and women, with 1.6 million deaths. For men, TBL cancer was the leading cause of DALYs (24.9 million). For women, breast cancer was the leading cause of DALYs (13.1 million). Age-standardized incidence rates (ASIRs) per 100 000 and age-standardized death rates (ASDRs) per 100 000 for both sexes in 2013 were higher in developing vs developed countries for stomach cancer (ASIR, 17 vs 14; ASDR, 15 vs 11), liver cancer (ASIR, 15 vs 7; ASDR, 16 vs 7), esophageal cancer (ASIR, 9 vs 4; ASDR, 9 vs 4), cervical cancer (ASIR, 8 vs 5; ASDR, 4 vs 2), lip and oral cavity cancer (ASIR, 7 vs 6; ASDR, 2 vs 2), and nasopharyngeal cancer (ASIR, 1.5 vs 0.4; ASDR, 1.2 vs 0.3). Between 1990 and 2013, ASIRs for all cancers combined (except nonmelanoma skin cancer and Kaposi sarcoma) increased by more than 10% in 113 countries and decreased by more than 10% in 12 of 188 countries. Conclusions and Relevance Cancer poses a major threat to public health worldwide, and incidence rates have increased in most countries since 1990. The trend is a particular threat to developing nations with health systems that are ill-equipped to deal with complex and expensive cancer treatments. The annual update on the Global Burden of Cancer will provide all stakeholders with timely estimates to guide policy efforts in cancer prevention, screening, treatment, and palliation.

Nivolumab and ipilimumab versus ipilimumab in untreated melanoma

Abstract: BackgroundIn a phase 1 dose-escalation study, combined inhibition of T-cell checkpoint pathways by nivolumab and ipilimumab was associated with a high rate of objective response, including complete responses, among patients with advanced melanoma. MethodsIn this double-blind study involving 142 patients with metastatic melanoma who had not previously received treatment, we randomly assigned patients in a 2:1 ratio to receive ipilimumab (3 mg per kilogram of body weight) combined with either nivolumab (1 mg per kilogram) or placebo once every 3 weeks for four doses, followed by nivolumab (3 mg per kilogram) or placebo every 2 weeks until the occurrence of disease progression or unacceptable toxic effects. The primary end point was the rate of investigator-assessed, confirmed objective response among patients with BRAF V600 wild-type tumors. ResultsAmong patients with BRAF wild-type tumors, the rate of confirmed objective response was 61% (44 of 72 patients) in the group that received both ipilimumab and ni...

Human-level concept learning through probabilistic program induction.

Abstract: People learning new concepts can often generalize successfully from just a single example, yet machine learning algorithms typically require tens or hundreds of examples to perform with similar accuracy. People can also use learned concepts in richer ways than conventional algorithms-for action, imagination, and explanation. We present a computational model that captures these human learning abilities for a large class of simple visual concepts: handwritten characters from the world's alphabets. The model represents concepts as simple programs that best explain observed examples under a Bayesian criterion. On a challenging one-shot classification task, the model achieves human-level performance while outperforming recent deep learning approaches. We also present several "visual Turing tests" probing the model's creative generalization abilities, which in many cases are indistinguishable from human behavior.

Anticancer immunotherapy by CTLA-4 blockade relies on the gut microbiota

Abstract: Antibodies targeting CTLA-4 have been successfully used as cancer immunotherapy. We find that the antitumor effects of CTLA-4 blockade depend on distinct Bacteroides species. In mice and patients, T cell responses specific for B. thetaiotaomicron or B. fragilis were associated with the efficacy of CTLA-4 blockade. Tumors in antibiotic-treated or germ-free mice did not respond to CTLA blockade. This defect was overcome by gavage with B. fragilis, by immunization with B. fragilis polysaccharides, or by adoptive transfer of B. fragilis–specific T cells. Fecal microbial transplantation from humans to mice confirmed that treatment of melanoma patients with antibodies against CTLA-4 favored the outgrowth of B. fragilis with anticancer properties. This study reveals a key role for Bacteroidales in the immunostimulatory effects of CTLA-4 blockade.

Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-scale Convolutional Architecture

Abstract: In this paper we address three different computer vision tasks using a single basic architecture: depth prediction, surface normal estimation, and semantic labeling. We use a multiscale convolutional network that is able to adapt easily to each task using only small modifications, regressing from the input image to the output map directly. Our method progressively refines predictions using a sequence of scales, and captures many image details without any superpixels or low-level segmentation. We achieve state-of-the-art performance on benchmarks for all three tasks.

Character-level Convolutional Networks for Text Classification

Abstract: This article offers an empirical exploration on the use of character-level convolutional networks (ConvNets) for text classification. We constructed several large-scale datasets to show that character-level convolutional networks could achieve state-of-the-art or competitive results. Comparisons are offered against traditional models such as bag of words, n-grams and their TFIDF variants, and deep learning models such as word-based ConvNets and recurrent neural networks.

Deep generative image models using a Laplacian pyramid of adversarial networks

Abstract: In this paper we introduce a generative parametric model capable of producing high quality samples of natural images. Our approach uses a cascade of convolutional networks within a Laplacian pyramid framework to generate images in a coarse-to-fine fashion. At each level of the pyramid, a separate generative convnet model is trained using the Generative Adversarial Nets (GAN) approach [11]. Samples drawn from our model are of significantly higher quality than alternate approaches. In a quantitative assessment by human evaluators, our CIFAR10 samples were mistaken for real images around 40% of the time, compared to 10% for samples drawn from a GAN baseline model. We also show samples from models trained on the higher resolution images of the LSUN scene dataset.

End-to-end memory networks

Abstract: We introduce a neural network with a recurrent attention model over a possibly large external memory. The architecture is a form of Memory Network [23] but unlike the model in that work, it is trained end-to-end, and hence requires significantly less supervision during training, making it more generally applicable in realistic settings. It can also be seen as an extension of RNNsearch [2] to the case where multiple computational steps (hops) are performed per output symbol. The flexibility of the model allows us to apply it to tasks as diverse as (synthetic) question answering [22] and to language modeling. For the former our approach is competitive with Memory Networks, but with less supervision. For the latter, on the Penn TreeBank and Text8 datasets our approach demonstrates comparable performance to RNNs and LSTMs. In both cases we show that the key concept of multiple computational hops yields improved results.

librosa: Audio and Music Signal Analysis in Python

Abstract: This document describes version 0.4.0 of librosa: a Python pack- age for audio and music signal processing. At a high level, librosa provides implementations of a variety of common functions used throughout the field of music information retrieval. In this document, a brief overview of the library's functionality is provided, along with explanations of the design goals, software development practices, and notational conventions.

Combined Measurement of the Higgs Boson Mass in pp Collisions at √s=7 and 8 TeV with the ATLAS and CMS Experiments

Abstract: A measurement of the Higgs boson mass is presented based on the combined data samples of the ATLAS and CMS experiments at the CERN LHC in the H→γγ and H→ZZ→4l decay channels. The results are obtained from a simultaneous fit to the reconstructed invariant mass peaks in the two channels and for the two experiments. The measured masses from the individual channels and the two experiments are found to be consistent among themselves. The combined measured mass of the Higgs boson is mH=125.09±0.21 (stat)±0.11 (syst) GeV.

An Empirical Exploration of Recurrent Network Architectures

Abstract: The Recurrent Neural Network (RNN) is an extremely powerful sequence model that is often difficult to train. The Long Short-Term Memory (LSTM) is a specific RNN architecture whose design makes it much easier to train. While wildly successful in practice, the LSTM's architecture appears to be ad-hoc so it is not clear if it is optimal, and the significance of its individual components is unclear. In this work, we aim to determine whether the LSTM architecture is optimal or whether much better architectures exist. We conducted a thorough architecture search where we evaluated over ten thousand different RNN architectures, and identified an architecture that outperforms both the LSTM and the recently-introduced Gated Recurrent Unit (GRU) on some but not all tasks. We found that adding a bias of 1 to the LSTM's forget gate closes the gap between the LSTM and the GRU.

Pembrolizumab versus investigator-choice chemotherapy for ipilimumab-refractory melanoma (KEYNOTE-002): a randomised, controlled, phase 2 trial

Abstract: Summary Background Patients with melanoma that progresses on ipilimumab and, if BRAF V600 mutant-positive, a BRAF or MEK inhibitor or both, have few treatment options. We assessed the efficacy and safety of two pembrolizumab doses versus investigator-choice chemotherapy in patients with ipilimumab-refractory melanoma. Methods We carried out a randomised phase 2 trial of patients aged 18 years or older from 73 hospitals, clinics, and academic medical centres in 12 countries who had confirmed progressive disease within 24 weeks after two or more ipilimumab doses and, if BRAF V600 mutant-positive, previous treatment with a BRAF or MEK inhibitor or both. Patients had to have resolution of all ipilimumab-related adverse events to grade 0–1 and prednisone 10 mg/day or less for at least 2 weeks, an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1, and at least one measurable lesion to be eligible. Using a centralised interactive voice response system, we randomly assigned (1:1:1) patients in a block size of six to receive intravenous pembrolizumab 2 mg/kg or 10 mg/kg every 3 weeks or investigator-choice chemotherapy (paclitaxel plus carboplatin, paclitaxel, carboplatin, dacarbazine, or oral temozolomide). Randomisation was stratified by ECOG performance status, lactate dehydrogenase concentration, and BRAF V600 mutation status. Individual treatment assignment between pembrolizumab and chemotherapy was open label, but investigators and patients were masked to assignment of the dose of pembrolizumab. We present the primary endpoint at the prespecified second interim analysis of progression-free survival in the intention-to-treat population. This study is registered with ClinicalTrials.gov, number NCT01704287. The study is closed to enrolment but continues to follow up and treat patients. Findings Between Nov 30, 2012, and Nov 13, 2013, we enrolled 540 patients: 180 patients were randomly assigned to receive pembrolizumab 2 mg/kg, 181 to receive pembrolizumab 10 mg/kg, and 179 to receive chemotherapy. Based on 410 progression-free survival events, progression-free survival was improved in patients assigned to pembrolizumab 2 mg/kg (HR 0·57, 95% CI 0·45–0·73; p Interpretation These findings establish pembrolizumab as a new standard of care for the treatment of ipilimumab-refractory melanoma. Funding Merck Sharp & Dohme.

End-To-End Memory Networks

Abstract: We introduce a neural network with a recurrent attention model over a possibly large external memory. The architecture is a form of Memory Network (Weston et al., 2015) but unlike the model in that work, it is trained end-to-end, and hence requires significantly less supervision during training, making it more generally applicable in realistic settings. It can also be seen as an extension of RNNsearch to the case where multiple computational steps (hops) are performed per output symbol. The flexibility of the model allows us to apply it to tasks as diverse as (synthetic) question answering and to language modeling. For the former our approach is competitive with Memory Networks, but with less supervision. For the latter, on the Penn TreeBank and Text8 datasets our approach demonstrates comparable performance to RNNs and LSTMs. In both cases we show that the key concept of multiple computational hops yields improved results.

Deep multi-scale video prediction beyond mean square error

Abstract: Learning to predict future images from a video sequence involves the construction of an internal representation that models the image evolution accurately, and therefore, to some degree, its content and dynamics. This is why pixel-space video prediction may be viewed as a promising avenue for unsupervised feature learning. In addition, while optical flow has been a very studied problem in computer vision for a long time, future frame prediction is rarely approached. Still, many vision applications could benefit from the knowledge of the next frames of videos, that does not require the complexity of tracking every pixel trajectories. In this work, we train a convolutional network to generate future frames given an input sequence. To deal with the inherently blurry predictions obtained from the standard Mean Squared Error (MSE) loss function, we propose three different and complementary feature learning strategies: a multi-scale architecture, an adversarial training method, and an image gradient difference loss function. We compare our predictions to different published results based on recurrent neural networks on the UCF101 dataset

Programmable materials and the nature of the DNA bond

Abstract: For over half a century, the biological roles of nucleic acids as catalytic enzymes, intracellular regulatory molecules, and the carriers of genetic information have been studied extensively. More recently, the sequence-specific binding properties of DNA have been exploited to direct the assembly of materials at the nanoscale. Integral to any methodology focused on assembling matter from smaller pieces is the idea that final structures have well-defined spacings, orientations, and stereo-relationships. This requirement can be met by using DNA-based constructs that present oriented nanoscale bonding elements from rigid core units. Here, we draw analogy between such building blocks and the familiar chemical concepts of "bonds" and "valency" and review two distinct but related strategies that have used this design principle in constructing new configurations of matter.

Clearance systems in the brain—implications for Alzheimer disease

Abstract: Accumulation of toxic protein aggregates-amyloid-β (Aβ) plaques and hyperphosphorylated tau tangles-is the pathological hallmark of Alzheimer disease (AD). Aβ accumulation has been hypothesized to result from an imbalance between Aβ production and clearance; indeed, Aβ clearance seems to be impaired in both early and late forms of AD. To develop efficient strategies to slow down or halt AD, it is critical to understand how Aβ is cleared from the brain. Extracellular Aβ deposits can be removed from the brain by various clearance systems, most importantly, transport across the blood-brain barrier. Findings from the past few years suggest that astroglial-mediated interstitial fluid (ISF) bulk flow, known as the glymphatic system, might contribute to a larger portion of extracellular Aβ (eAβ) clearance than previously thought. The meningeal lymphatic vessels, discovered in 2015, might provide another clearance route. Because these clearance systems act together to drive eAβ from the brain, any alteration to their function could contribute to AD. An understanding of Aβ clearance might provide strategies to reduce excess Aβ deposits and delay, or even prevent, disease onset. In this Review, we describe the clearance systems of the brain as they relate to proteins implicated in AD pathology, with the main focus on Aβ.