Genetic linkage maps play fundamental roles in understanding genome structure, explaining genome formation events during evolution, and discovering the genetic bases of important traits. A high-density cotton (Gossypium spp.) genetic map was developed using representative sets of simple sequence repeat (SSR) and the first public set of single nucleotide polymorphism (SNP) markers to genotype 186 recombinant inbred lines (RILs) derived from an interspecific cross between Gossypium hirsutum L. (TM-1) and G. barbadense L. (3-79). The genetic map comprised 2072 loci (1825 SSRs and 247 SNPs) and covered 3380 centiMorgan (cM) of the cotton genome (AD) with an average marker interval of 1.63 cM. The allotetraploid cotton genome produced equivalent recombination frequencies in its two subgenomes (At and Dt). Of the 2072 loci, 1138 (54.9%) were mapped to 13 At-subgenome chromosomes, covering 1726.8 cM (51.1%), and 934 (45.1%) mapped to 13 Dt-subgenome chromosomes, covering 1653.1 cM (48.9%). The genetically smallest homeologous chromosome pair was Chr. 04 (A04) and 22 (D04), and the largest was Chr. 05 (A05) and 19 (D05). Duplicate loci between and within homeologous chromosomes were identified that facilitate investigations of chromosome translocations. The map augments evidence of reciprocal rearrangement between ancestral forms of Chr. 02 and 03 versus segmental homeologs 14 and 17 as centromeric regions show homeologous between Chr. 02 (A02) and 17 (D02), as well as between Chr. 03 (A03) and 14 (D03). This research represents an important foundation for studies on polyploid cottons, including germplasm characterization, gene discovery, and genome sequence assembly.

Variational autoencoders (VAE) combined with hierarchical RNNs have emerged as a powerful framework for conversation modeling. However, they suffer from the notorious degeneration problem, where the decoders learn to ignore latent variables and reduce to vanilla RNNs. We empirically show that this degeneracy occurs mostly due to two reasons. First, the expressive power of hierarchical RNN decoders is often high enough to model the data using only its decoding distributions without relying on the latent variables. Second, the conditional VAE structure whose generation process is conditioned on a context, makes the range of training targets very sparse; that is, the RNN decoders can easily overfit to the training data ignoring the latent variables. To solve the degeneration problem, we propose a novel model named Variational Hierarchical Conversation RNNs (VHCR), involving two key ideas of (1) using a hierarchical structure of latent variables, and (2) exploiting an utterance drop regularization. With evaluations on two datasets of Cornell Movie Dialog and Ubuntu Dialog Corpus, we show that our VHCR successfully utilizes latent variables and outperforms state-of-the-art models for conversation generation. Moreover, it can perform several new utterance control tasks, thanks to its hierarchical latent structure.

No abstract

High signal on T1 -weighted image and numerous signal voids are highly suggestive of ASPS, although they are not universal as has been suggested and arteriovenous malformation should be included in the differential diagnosis. Local bony metastases in ASPS were seen in two cases and should be carefully investigated.

Mirroring the success of masked language models, vision-and-language counterparts like VILBERT, LXMERT and UNITER have achieved state of the art performance on a variety of multimodal discriminative tasks like visual question answering and visual grounding. Recent work has also successfully adapted such models towards the generative task of image captioning. This begs the question: Can these models go the other way and generate images from pieces of text? Our analysis of a popular representative from this model family -LXMERT -finds that it is unable to generate rich and semantically meaningful imagery with its current training setup. We introduce X-LXMERT, an extension to LXMERT with training refinements including: discretizing visual representations, using uniform masking with a large range of masking ratios and aligning the right pre-training datasets to the right objectives which enables it to paint. X-LXMERT's image generation capabilities rival state of the art generative models while its question answering and captioning abilities remains comparable to LXMERT. Finally, we demonstrate the generality of these training refinements by adding image generation capabilities into UNITER to produce X-UNITER.

Generating diverse sequences is important in many NLP applications such as question generation or summarization that exhibit semantically one-to-many relationships between source and the target sequences. We present a method to explicitly separate diversification from generation using a general plug-and-play module (called SELECTOR) that wraps around and guides an existing encoder-decoder model. The diversification stage uses a mixture of experts to sample different binary masks on the source sequence for diverse content selection. The generation stage uses a standard encoder-decoder model given each selected content from the source sequence. Due to the non-differentiable nature of discrete sampling and the lack of ground truth labels for binary mask, we leverage a proxy for ground-truth mask and adopt stochastic hard-EM for training. In question generation (SQuAD) and abstractive summarization (CNN-DM), our method demonstrates significant improvements in accuracy, diversity and training efficiency, including state-of-the-art top-1 accuracy in both datasets, 6% gain in top-5 accuracy, and 3.7 times faster training over a state-of-the-art model. Our code is publicly available at https://github.com/ clovaai/FocusSeq2Seq.

BackgroundUpland cotton (G. hirsutum L.) is the leading fiber crop worldwide. Genetic improvement of fiber quality and yield is facilitated by a variety of genomics tools. An integrated genetic and physical map is needed to better characterize quantitative trait loci and to allow for the positional cloning of valuable genes. However, developing integrated genomic tools for complex allotetraploid genomes, like that of cotton, is highly experimental. In this report, we describe an effective approach for developing an integrated physical framework that allows for the distinguishing between subgenomes in cotton.ResultsA physical map has been developed with 220 and 115 BAC contigs for homeologous chromosomes 12 and 26, respectively, covering 73.49 Mb and 34.23 Mb in physical length. Approximately one half of the 220 contigs were anchored to the At subgenome only, while 48 of the 115 contigs were allocated to the Dt subgenome only. Between the two chromosomes, 67 contigs were shared with an estimated overall physical similarity between the two chromosomal homeologs at 40.0 %. A total of 401 fiber unigenes plus 214 non-fiber unigenes were located to chromosome 12 while 207 fiber unigenes plus 183 non-fiber unigenes were allocated to chromosome 26. Anchoring was done through an overgo hybridization approach and all anchored ESTs were functionally annotated via blast analysis.ConclusionThis integrated genomic map describes the first pair of homoeologous chromosomes of an allotetraploid genome in which BAC contigs were identified and partially separated through the use of chromosome-specific probes and locus-specific genetic markers. The approach used in this study should prove useful in the construction of genome-wide physical maps for polyploid plant genomes including Upland cotton. The identification of Gene-rich islands in the integrated map provides a platform for positional cloning of important genes and the targeted sequencing of specific genomic regions.

Because of the facile formation of defects in organometal halide perovskites, the defect passivation has become an important prerequisite for the stable and efficient perovskite solar cell (PSC). Regarding that ionic defects of the perovskites play a significant role on the performance and stability of PSCs, we introduce lithium fluorides as effective passivators based on their strong ionic characteristics and small ionic radii. Both Li+ and F– are observed to successfully incorporate within the perovskite layer, improving the device performances with the best efficiency over 20%, while the hysteresis effects are significantly reduced, confirming the passivation of perovskite defects. Moreover, LiF restrains both thermal degradation and photodegradation of PSCs, where over 90% of the initial efficiencies have been retained by LiF-incorporated devices for more than 1000 h under either 1 sun illumination or 85 °C thermal condition. As the trap density of states is analyzed before and after the thermal stress, not only the mitigation of electronic traps as fabricated but also the dramatic relaxation of traps during the postannealing step is observed with the LiF incorporation. From this work, LiF has shown its potential as a promising ionic passivator, and the phenomenal achievement of device stability by LiF provides a clear insight to overcome the stability issues of PSCs, a key to the commercialization of next-generation photovoltaics.