No abstract

Image captioning is an important but challenging task, applicable to virtual assistants, editing tools, image indexing, and support of the disabled. Its challenges are due to the variability and ambiguity of possible image descriptions. In recent years significant progress has been made in image captioning, using Recurrent Neural Networks powered by long-short-term-memory (LSTM) units. Despite mitigating the vanishing gradient problem, and despite their compelling ability to memorize dependencies, LSTM units are complex and inherently sequential across time. To address this issue, recent work has shown benefits of convolutional networks for machine translation and conditional image generation [9,34,35]. Inspired by their success, in this paper, we develop a convolutional image captioning technique. We demonstrate its efficacy on the challenging MSCOCO dataset and demonstrate performance on par with the baseline, while having a faster training time per number of parameters. We also perform a detailed analysis, providing compelling reasons in favor of convolutional language generation approaches.

No abstract

Colorization is an ambiguous problem, with multiple viable colorizations for a single grey-level image. However, previous methods only produce the single most probable colorization. Our goal is to model the diversity intrinsic to the problem of colorization and produce multiple colorizations that display long-scale spatial co-ordination. We learn a low dimensional embedding of color fields using a variational autoencoder (VAE). We construct loss terms for the VAE decoder that avoid blurry outputs and take into account the uneven distribution of pixel colors. Finally, we build a conditional model for the multi-modal distribution between grey-level image and the color field embeddings. Samples from this conditional model result in diverse colorization. We demonstrate that our method obtains better diverse colorizations than a standard conditional variational autoencoder (CVAE) model, as well as a recently proposed conditional generative adversarial network (cGAN).

Image captioning is an ambiguous problem, with many suitable captions for an image. To address ambiguity, beam search is the de facto method for sampling multiple captions. However, beam search is computationally expensive and known to produce generic captions [8,10]. To address this concern, some variational auto-encoder (VAE) [32] and generative adversarial net (GAN) [5,25] based methods have been proposed. Though diverse, GAN and VAE are less accurate. In this paper, we first predict a meaningful summary of the image, then generate the caption based on that summary. We use part-of-speech as summaries, since our summary should drive caption generation. We achieve the trifecta: (1) High accuracy for the diverse captions as evaluated by standard captioning metrics and user studies; (2) Faster computation of diverse captions compared to beam search and diverse beam search [28]; and (3) High diversity as evaluated by counting novel sentences, distinct n-grams and mutual overlap (i.e., mBleu-4) scores.

No abstract

There is evidence that certain club cells (CCs) in the murine airways associated with neuroepithelial bodies (NEBs) and terminal bronchioles are resistant to the xenobiotic naphthalene (Nap) and repopulate the airways after Nap injury. The identity and significance of these progenitors (variant CCs, v-CCs) have remained elusive. A recent screen for CC markers identified rare Uroplakin3a (Upk3a)-expressing cells (U-CCs) with a v-CC-like distribution. Here, we employ lineage analysis in the uninjured and chemically injured lungs to investigate the role of U-CCs as epithelial progenitors. U-CCs proliferate and generate CCs and ciliated cells in uninjured airways long-term and, like v-CCs, after Nap. U-CCs have a higher propensity to generate ciliated cells than non-U-CCs. Although U-CCs do not contribute to alveolar maintenance long-term, they generate alveolar type I and type II cells after Bleomycin (Bleo)-induced alveolar injury. Finally, we report that Upk3a cells exist in the NEB microenvironment of the human lung and are aberrantly expanded in conditions associated with neuroendocrine hyperplasias.

Uterine leiomyosarcomas (uLMS) are aggressive tumors arising from the smooth muscle layer of the uterus. We analyzed 83 uLMS sample genetics, including 56 from Yale and 27 from The Cancer Genome Atlas (TCGA). Among them, a total of 55 Yale samples including two patient-derived xenografts (PDXs) and 27 TCGA samples have whole-exome sequencing (WES) data; 10 Yale and 27 TCGA samples have RNA-sequencing (RNA-Seq) data; and 11 Yale and 10 TCGA samples have whole-genome sequencing (WGS) data. We found recurrent somatic mutations in TP53, MED12, and PTEN genes. Top somatic mutated genes included TP53, ATRX, PTEN, and MEN1 genes. Somatic copy number variation (CNV) analysis identified 8 copy-number gains, including 5p15.33 (TERT), 8q24.21 (C-MYC), and 17p11.2 (MYOCD, MAP2K4) amplifications and 29 copy-number losses. Fusions involving tumor suppressors or oncogenes were deetected, with most fusions disrupting RB1, TP53, and ATRX/DAXX, and one fusion (ACTG2-ALK) being potentially targetable. WGS results demonstrated that 76% (16 of 21) of the samples harbored chromoplexy and/or chromothripsis. Clinically actionable mutational signatures of homologous-recombination DNA-repair deficiency (HRD) and microsatellite instability (MSI) were identified in 25% (12 of 48) and 2% (1 of 48) of fresh frozen uLMS, respectively. Finally, we found olaparib (PARPi; P = 0.002), GS-626510 (C-MYC/BETi; P < 0.000001 and P = 0.0005), and copanlisib (PIK3CAi; P = 0.0001) monotherapy to significantly inhibit uLMS-PDXs harboring derangements in C-MYC and PTEN/PIK3CA/AKT genes (LEY11) and/or HRD signatures (LEY16) compared to vehicle-treated mice. These findings define the genetic landscape of uLMS and suggest that a subset of uLMS may benefit from existing PARP-, PIK3CA-, and C-MYC/BET-targeted drugs.