scispace - formally typeset
Search or ask a question

Showing papers by "Google published in 2022"


Journal ArticleDOI
TL;DR: For example, this article used deep learning models to predict functional annotations for unaligned amino acid sequences across rigorous benchmark assessments built from the 17,929 families of the protein families database Pfam.
Abstract: Understanding the relationship between amino acid sequence and protein function is a long-standing challenge with far-reaching scientific and translational implications. State-of-the-art alignment-based techniques cannot predict function for one-third of microbial protein sequences, hampering our ability to exploit data from diverse organisms. Here, we train deep learning models to accurately predict functional annotations for unaligned amino acid sequences across rigorous benchmark assessments built from the 17,929 families of the protein families database Pfam. The models infer known patterns of evolutionary substitutions and learn representations that accurately cluster sequences from unseen families. Combining deep models with existing methods significantly improves remote homology detection, suggesting that the deep models learn complementary information. This approach extends the coverage of Pfam by >9.5%, exceeding additions made over the last decade, and predicts function for 360 human reference proteome proteins with no previous Pfam annotation. These results suggest that deep learning models will be a core component of future protein annotation tools. A deep learning model predicts protein functional annotations for unaligned amino acid sequences.

90 citations


Journal ArticleDOI
TL;DR: In this article , the authors used Long Short-Term Memory (LSTM) networks and an LSTM variant that is architecturally constrained to conserve mass to predict extreme events.
Abstract: Abstract. The most accurate rainfall–runoff predictions are currently based on deep learning. There is a concern among hydrologists that the predictive accuracy of data-driven models based on deep learning may not be reliable in extrapolation or for predicting extreme events. This study tests that hypothesis using long short-term memory (LSTM) networks and an LSTM variant that is architecturally constrained to conserve mass. The LSTM network (and the mass-conserving LSTM variant) remained relatively accurate in predicting extreme (high-return-period) events compared with both a conceptual model (the Sacramento Model) and a process-based model (the US National Water Model), even when extreme events were not included in the training period. Adding mass balance constraints to the data-driven model (LSTM) reduced model skill during extreme events.

21 citations


Journal ArticleDOI
TL;DR: In this article , a simple regression approach was used to map the LSTM state vector to the target stores (soil moisture and snow) of interest, and good correlations between the probe outputs and the target variables of interest provided evidence that LSTMs contain information that reflects known hydrological processes comparable with the concept of variable-capacity soil moisture stores.
Abstract: Abstract. Neural networks have been shown to be extremely effective rainfall-runoff models, where the river discharge is predicted from meteorological inputs. However, the question remains: what have these models learned? Is it possible to extract information about the learned relationships that map inputs to outputs, and do these mappings represent known hydrological concepts? Small-scale experiments have demonstrated that the internal states of long short-term memory networks (LSTMs), a particular neural network architecture predisposed to hydrological modelling, can be interpreted. By extracting the tensors which represent the learned translation from inputs (precipitation, temperature, and potential evapotranspiration) to outputs (discharge), this research seeks to understand what information the LSTM captures about the hydrological system. We assess the hypothesis that the LSTM replicates real-world processes and that we can extract information about these processes from the internal states of the LSTM. We examine the cell-state vector, which represents the memory of the LSTM, and explore the ways in which the LSTM learns to reproduce stores of water, such as soil moisture and snow cover. We use a simple regression approach to map the LSTM state vector to our target stores (soil moisture and snow). Good correlations (R2>0.8) between the probe outputs and the target variables of interest provide evidence that the LSTM contains information that reflects known hydrological processes comparable with the concept of variable-capacity soil moisture stores. The implications of this study are threefold: (1) LSTMs reproduce known hydrological processes. (2) While conceptual models have theoretical assumptions embedded in the model a priori, the LSTM derives these from the data. These learned representations are interpretable by scientists. (3) LSTMs can be used to gain an estimate of intermediate stores of water such as soil moisture. While machine learning interpretability is still a nascent field and our approach reflects a simple technique for exploring what the model has learned, the results are robust to different initial conditions and to a variety of benchmarking experiments. We therefore argue that deep learning approaches can be used to advance our scientific goals as well as our predictive goals.

19 citations


Journal ArticleDOI
Bara Barakat1
TL;DR: In this paper , the authors performed a systematic search using multiple databases (PubMed, MEDLINE, EMBASE, and Cochrane Central) until March 2021 to evaluate the effectiveness of Retzius sparing robot-assisted radical prostatectomy (RS-RARP) compared with standard RARP, in terms of perioperative, functional, and oncological outcomes.
Abstract: Retzius sparing robot-assisted radical prostatectomy (RS-RARP) is increasingly being used, but results of pertinent studies on perioperative, functional, and oncological outcomes comparing the Retzius sparing approach with standard robot-assisted radical prostatectomy (RARP) remain inconsistent.To evaluate the effectiveness of RS-RARP compared with standard RARP, in terms of perioperative, functional, and oncological outcomes.We performed a systematic search using multiple databases (PubMed, MEDLINE, EMBASE, and Cochrane Central) until March 2021. Only randomized controlled trials (RCTs) and prospective studies were eligible for study inclusion. Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines were respected. Studies were critically appraised for the risk of bias. Primary outcomes were continence/potency recovery, as well as positive surgical margin (PSM) rates. Secondary outcomes included total intra- and perioperative complication rates.Four RCTs and six prospective observational studies were included in this systematic review. The meta-analysis revealed that PSM rates in ≤pT2 tumors were statistically significantly higher, following RS-RARP as compared with RARP (risk ratio [RR]=1.39; 95% confidence interval [CI]=[1.01-1.91]). PSM rates in ≥pT3 tumors tended to be higher following RS-RARP (RR=1.36; 95% CI=[0.74-2.50]), although statistical significance was not reached. Immediate continence recovery was higher and significantly advantageous for RS-RARP (RR=1.81; 95% CI=[1.26-2.60]). Continence recovery also tended to be higher at 3 and 6 mo in the RS-RARP group (RR=1.57; 95% CI=[0.69-3.58] and RR=1.22; 95% CI=[0.89-1.66], respectively). The urinary continence recovery at 12 mo was similar in both groups (RR=1.14; 95% CI=[0.98-1.32]). A meta-analysis of included studies showed no significant difference concerning the return of erectile function and major complication rates between RS-RARP and RARP (RR=1.05; 95% CI=[0.76-1.45] and (RR=0.79; 95% CI=[0.07-8.74], respectively).Available data suggest a statistically significant advantage in favor of RS-RARP in terms of immediate urinary continence recovery. PSM rates in localized ≤pT2 tumors are statistically significantly higher following RS-RARP. Potency and serious complication rates appear to be similar.Our meta-analysis of the current evidence shows a significant advantage for Retzius sparing robot-assisted radical prostatectomy (RS-RARP) over robot-assisted radical prostatectomy in terms of immediate urinary continence recovery, but positive cancer margins are higher following RS-RARP. There was no significant difference in the preservation of erectile function and overall postoperative complication rates between both the techniques.

17 citations


Journal ArticleDOI
TL;DR: Google's operational flood forecasting system was developed to provide accurate real-time flood warnings to agencies and the public with a focus on riverine floods in large, gauged rivers as discussed by the authors .
Abstract: Abstract. Google's operational flood forecasting system was developed to provide accurate real-time flood warnings to agencies and the public with a focus on riverine floods in large, gauged rivers. It became operational in 2018 and has since expanded geographically. This forecasting system consists of four subsystems: data validation, stage forecasting, inundation modeling, and alert distribution. Machine learning is used for two of the subsystems. Stage forecasting is modeled with the long short-term memory (LSTM) networks and the linear models. Flood inundation is computed with the thresholding and the manifold models, where the former computes inundation extent and the latter computes both inundation extent and depth. The manifold model, presented here for the first time, provides a machine-learning alternative to hydraulic modeling of flood inundation. When evaluated on historical data, all models achieve sufficiently high-performance metrics for operational use. The LSTM showed higher skills than the linear model, while the thresholding and manifold models achieved similar performance metrics for modeling inundation extent. During the 2021 monsoon season, the flood warning system was operational in India and Bangladesh, covering flood-prone regions around rivers with a total area close to 470 000 km2, home to more than 350 000 000 people. More than 100 000 000 flood alerts were sent to affected populations, to relevant authorities, and to emergency organizations. Current and future work on the system includes extending coverage to additional flood-prone locations and improving modeling capabilities and accuracy.

14 citations


Journal ArticleDOI
TL;DR: In this article , a computational method to identify functionally important transcription factors (TFs) from chromatin accessibility measurements is presented, which applies an ontology-guided functional approach to compute novel enrichment by integrating accessibility measurements, high-confidence pre-computed conservation-aware TF binding sites, and putative gene-regulatory models.
Abstract: We present WhichTF, a computational method to identify functionally important transcription factors (TFs) from chromatin accessibility measurements. To rank TFs, WhichTF applies an ontology-guided functional approach to compute novel enrichment by integrating accessibility measurements, high-confidence pre-computed conservation-aware TF binding sites, and putative gene-regulatory models. Comparison with prior sheer abundance-based methods reveals the unique ability of WhichTF to identify context-specific TFs with functional relevance, including NF-κB family members in lymphocytes and GATA factors in cardiac cells. To distinguish the transcriptional regulatory landscape in closely related samples, we apply differential analysis and demonstrate its utility in lymphocyte, mesoderm developmental, and disease cells. We find suggestive, under-characterized TFs, such as RUNX3 in mesoderm development and GLI1 in systemic lupus erythematosus. We also find TFs known for stress response, suggesting routine experimental caveats that warrant careful consideration. WhichTF yields biological insight into known and novel molecular mechanisms of TF-mediated transcriptional regulation in diverse contexts, including human and mouse cell types, cell fate trajectories, and disease-associated cells.

11 citations


Journal ArticleDOI
TL;DR: In this article , a mixed-integer-programming-based optimization approach was proposed to maximize the fabrication yield of fixed-frequency superconducting quantum processors with cross-resonance interaction processors.
Abstract: Fixed-frequency superconducting quantum processors are one of the most mature quantum computing architectures with high-coherence qubits and simple controls. However, high-fidelity multi-qubit gates pose tight requirements on individual qubit frequencies in these processors , and these constraints are difficult to satisfy when constructing larger processors due to the large dispersion in the fabrication of Josephson junctions. In this article, we propose a mixed-integer-programming-based optimization approach that determines qubit frequencies to maximize the fabrication yield of quantum processors. We study traditional qubit and qutrit (three-level) architectures with cross-resonance interaction processors. We compare these architectures to a differential AC-Stark shift based on entanglement gates and show that our approach greatly improves the fabrication yield and also increases the scalability of these devices. Our approach is general and can be adapted to problems where one must avoid specific frequency collisions.

10 citations


Journal ArticleDOI
TL;DR: A family of functions is constructed suitable for establishing lower bounds on the oracle complexity of first-order minimization of smooth strongly-convex functions, and the new bounds match the known upper bounds up to a constant factor.

10 citations


Journal ArticleDOI
TL;DR: The first Continual Learning in Computer Vision Challenge (CLCVC) as discussed by the authors was held in 2019, which was one of the first opportunities to evaluate different continual learning algorithms on a common hardware with a large set of shared evaluation metrics and 3 different settings based on the realistic CORe50 video benchmark.

10 citations


Journal ArticleDOI
TL;DR: In this article , the authors presented an innovative proximity labeling (PL) strategy for single-cell-type proteomics of mouse brain, in which TurboID (an engineered biotin ligase) is used to label almost all proteins in a specific cell type.
Abstract: Proteome profiling is a powerful tool in biological and biomedical studies, starting with samples at bulk, single-cell, or single-cell-type levels. Reliable methods for extracting specific cell-type proteomes are in need, especially for the cells (e.g., neurons) that cannot be readily isolated. Here, we present an innovative proximity labeling (PL) strategy for single-cell-type proteomics of mouse brain, in which TurboID (an engineered biotin ligase) is used to label almost all proteins in a specific cell type. This strategy bypasses the requirement of cell isolation and includes five major steps: (i) constructing recombinant adeno-associated viruses (AAVs) to express TurboID driven by cell-type-specific promoters, (ii) delivering the AAV to mouse brains by direct intravenous injection, (iii) enhancing PL labeling by biotin administration, (iv) purifying biotinylated proteins, followed by on-bead protein digestion, and (v) quantitative tandem-mass-tag (TMT) labeling. We first confirmed that TurboID can label a wide range of cellular proteins in human HEK293 cells and optimized the single-cell-type proteomic pipeline. To analyze specific brain cell types, we generated recombinant AAVs to coexpress TurboID and mCherry proteins, driven by neuron- or astrocyte-specific promoters and validated the expected cell expression by coimmunostaining of mCherry and cellular markers. Subsequent biotin purification and TMT analysis identified ∼10,000 unique proteins from a few micrograms of protein samples with excellent reproducibility. Comparative and statistical analyses indicated that these PL proteomes contain cell-type-specific cellular pathways. Although PL was originally developed for studying protein–protein interactions and subcellular proteomes, we extended it to efficiently tag the entire proteomes of specific cell types in the mouse brain using TurboID biotin ligase. This simple, effective in vivo approach should be broadly applicable to single-cell-type proteomics.

10 citations


Journal ArticleDOI
TL;DR: In many applications of computer graphics, art, and design, it is desirable for a user to provide intuitive non-image input, such as text, sketch, stroke, graph or layout, and have a computer system automatically generate photo-realistic images according to that input as mentioned in this paper.
Abstract: In many applications of computer graphics, art, and design, it is desirable for a user to provide intuitive non-image input, such as text, sketch, stroke, graph, or layout, and have a computer system automatically generate photo-realistic images according to that input. While classically, works that allow such automatic image content generation have followed a framework of image retrieval and composition, recent advances in deep generative models such as generative adversarial networks (GANs), variational autoencoders (VAEs), and flow-based methods have enabled more powerful and versatile image generation approaches. This paper reviews recent works for image synthesis given intuitive user input, covering advances in input versatility, image generation methodology, benchmark datasets, and evaluation metrics. This motivates new perspectives on input representation and interactivity, cross fertilization between major image generation paradigms, and evaluation and comparison of generation methods.

Journal ArticleDOI
TL;DR: In this article, the authors present a CRDT algorithm that handles arbitrary concurrent modifications on trees, while ensuring that the tree structure remains valid (in particular, no cycles are introduced), and guaranteeing that all replicas converge towards the same consistent state.
Abstract: Replicated tree data structures are a fundamental building block of distributed filesystems, such as Google Drive and Dropbox, and collaborative applications with a JSON or XML data model. These systems need to support a move operation that allows a subtree to be moved to a new location within the tree. However, such a move operation is difficult to implement correctly if different replicas can concurrently perform arbitrary move operations, and we demonstrate bugs in Google Drive and Dropbox that arise with concurrent moves. In this article we present a CRDT algorithm that handles arbitrary concurrent modifications on trees, while ensuring that the tree structure remains valid (in particular, no cycles are introduced), and guaranteeing that all replicas converge towards the same consistent state. Our algorithm requires no synchronous coordination between replicas, making it highly available in the face of network partitions. We formally prove the correctness of our algorithm using the Isabelle/HOL proof assistant, and evaluate the performance of our formally verified implementation in a geo-replicated setting.

Posted Content
TL;DR: The authors proposed a new approach for paragraph identification by spatial graph convolutional neural networks (GCN) applied on OCR text boxes, where two steps, namely line splitting and line clustering, are performed to extract paragraphs from the lines in OCR results.
Abstract: Paragraphs are an important class of document entities. We propose a new approach for paragraph identification by spatial graph convolutional neural networks (GCN) applied on OCR text boxes. Two steps, namely line splitting and line clustering, are performed to extract paragraphs from the lines in OCR results. Each step uses a beta-skeleton graph constructed from bounding boxes, where the graph edges provide efficient support for graph convolution operations. With only pure layout input features, the GCN model size is 3~4 orders of magnitude smaller compared to R-CNN based models, while achieving comparable or better accuracies on PubLayNet and other datasets. Furthermore, the GCN models show good generalization from synthetic training data to real-world images, and good adaptivity for variable document styles.


Journal ArticleDOI
Blaise Aguera y Arcas1
01 Jan 2022-Daedalus
TL;DR: This article argued that large language models have a great deal to teach us about the nature of language, understanding, intelligence, sociality, and personhood, and that much of what we consider intelligence is inherently dialogic, hence social; it requires a theory of mind.
Abstract: Abstract Large language models (LLMs) represent a major advance in artificial intelligence and, in particular, toward the goal of human-like artificial general intelligence. It is sometimes claimed, though, that machine learning is “just statistics,” hence that, in this grander ambition, progress in AI is illusory. Here I take the contrary view that LLMs have a great deal to teach us about the nature of language, understanding, intelligence, sociality, and personhood. Specifically: statistics do amount to understanding, in any falsifiable sense. Furthermore, much of what we consider intelligence is inherently dialogic, hence social; it requires a theory of mind. Complex sequence learning and social interaction may be a sufficient basis for general intelligence, including theory of mind and consciousness. Since the interior state of another being can only be understood through interaction, no objective answer is possible to the question of when an “it” becomes a “who,” but for many people, neural nets running on computers are likely to cross this threshold in the very near future.

Posted ContentDOI
Aleksandra Machnik1
28 Mar 2022
TL;DR: In this article , the authors identify the barriers and factors to knowledge sharing within an emerging technology context and present a comprehensive in-depth interview with 38 professionals working in 5 organizations engaged in the field of emerging technologies to arrive at the findings discussed in this paper.
Abstract: As advances in emerging technologies continue to transform the business landscape, knowledge sharing will become increasingly important to leveraging the unique core competencies of organizations so as to gain a competitive advantage. Despite emerging technologies being so popular in news and media publications, knowledge sharing remains an area of research that is under-researched in the emerging technologies context. This research aims to identify the barriers and factors to knowledge sharing within an emerging technology context. A comprehensive in-depth interview was conducted with 38 professionals working in 5 organizations engaged in the field of emerging technologies in order to arrive at the findings discussed in this paper. Based on the analysis of the surveys, we found that there are six main factors driving the need for sharing knowledge. The six factors are: regular cadence, integrating expertise from different teams, diversity-inclusive social environment, Interconnected platforms that are accessible to all, a regular update schedule that needs to be followed, to create points of contact within different departments of an organization in order to facilitate sharing. It is important to note that the distribution and use of knowledge in organizations is dependent on the interactions between individuals.

Posted ContentDOI
Aleksandra Machnik1
23 Apr 2022
TL;DR: Among the essential elements of knowledge management is the use of information and data, as well as the knowledge, skills, and abilities inherent within communities, and their ideas, commitments, and motivations for making good decisions as emerging technologies become more prevalent as mentioned in this paper .
Abstract: Among the essential elements of knowledge management is the use of information and data, as well as the knowledge, skills, and abilities inherent within communities, as well as their ideas, commitments, and motivations for making good decisions as emerging technologies become more prevalent. Numerous leading social scientists in this field have asserted that organisational knowledge should be regarded as a strategic asset. There is a growing awareness of the importance of gathering, locating, capturing, and sharing collective knowledge and expertise of societies, and societies are urged to develop effective and efficient methods of gathering, locating, capturing, and sharing that knowledge in order to deal with problems and to benefit from opportunities. People living in many countries and regions are interested in implementing knowledge management processes and technologies, and many of them have included knowledge management as an integral part of their overall development strategies. The management of knowledge plays an increasingly important role in global economic development (Bell, 1973, 1978). In order to remain relevant in the modern world, organisations should not ignore knowledge management and emerging technologies.

Journal ArticleDOI
David Ha1, Stanley Devine1
01 Aug 2022
TL;DR: In this paper , the authors provide a historical context of neural network research's involvement with complex systems, and highlight several active areas in modern deep learning research that incorporate the principles of collective intelligence to advance its capabilities.
Abstract: In the past decade, we have witnessed the rise of deep learning to dominate the field of artificial intelligence. Advances in artificial neural networks alongside corresponding advances in hardware accelerators with large memory capacity, together with the availability of large datasets enabled practitioners to train and deploy sophisticated neural network models that achieve state-of-the-art performance on tasks across several fields spanning computer vision, natural language processing, and reinforcement learning. However, as these neural networks become bigger, more complex, and more widely used, fundamental problems with current deep learning models become more apparent. State-of-the-art deep learning models are known to suffer from issues that range from poor robustness, inability to adapt to novel task settings, to requiring rigid and inflexible configuration assumptions. Collective behavior, commonly observed in nature, tends to produce systems that are robust, adaptable, and have less rigid assumptions about the environment configuration. Collective intelligence, as a field, studies the group intelligence that emerges from the interactions of many individuals. Within this field, ideas such as self-organization, emergent behavior, swarm optimization, and cellular automata were developed to model and explain complex systems. It is therefore natural to see these ideas incorporated into newer deep learning methods. In this review, we will provide a historical context of neural network research’s involvement with complex systems, and highlight several active areas in modern deep learning research that incorporate the principles of collective intelligence to advance its capabilities. We hope this review can serve as a bridge between the complex systems and deep learning communities.

Journal ArticleDOI
Annalisa Pawlosky1
01 Jan 2022-iScience
TL;DR: In this article , the authors demonstrate early progress toward constructing a high-throughput, single-molecule protein sequencing technology utilizing barcoded DNA aptamers (binders) to recognize terminal amino acids of peptides (targets) tethered on a next-generation sequencing chip.

Posted ContentDOI
Aleksandra Machnik1
29 Mar 2022
TL;DR: In this paper , the authors examine the challenges and competitiveness that can be gained by digital governance on the world's business strategy using numerous real-world examples and anecdotes, and explain how concrete elements of digital governance can affect the competitiveness of companies.
Abstract: As a result of international competition and a globalized environment, companies today face the remarkable challenge of how to achieve and maintain their competitive advantage in a highly competitive environment in which international competitors play a significant role. The goal of digital governance has always been to make people's lives easier and improve customer satisfaction. In response to the recovery in markets, organizations around the world are now playing a different role. There is no doubt that companies are making considerable efforts to boost the factors which make them more competitive, including adoption of the principles of digital governance and automation. The role of digital governance is currently more of a facilitation than a regulation. Leaders and executives understand the importance of digital strategy and governance for competitive advantage. This paper examines the challenges and competitiveness that can be gained by digital governance on the world's business strategy using numerous real-world examples and anecdotes. This paper examines the challenges and competitiveness that can be gained by digital governance on the world's business strategy using numerous real- world examples and anecdotes. It is true that digitization has been around for some time, but there is still a long way to go to fully realize its potential. The aim of this paper is to explain how concrete elements of digital governance can affect the competitiveness of companies. Thus, digital governance was a key element in the successful recovery of the markets after the Covid era, despite the fact that it's been around for so long. In order for a company to be successful in international markets and maintain its quality, its productivity performance compared to its competitor’s digital governance plays a very important role.


Journal ArticleDOI
TL;DR: The authors propose a diagnostic dataset to test LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum, which can limit their utility, especially in the closed-book setting where the pretraining corpus must contain the facts the model should memorize.
Abstract: Many facts come with an expiration date, from the name of the President to the basketball team Lebron James plays for. But language models (LMs) are trained on snapshots of data collected at a specific moment in time, and this can limit their utility, especially in the closed-book setting where the pretraining corpus must contain the facts the model should memorize. We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum -- those trained on specific slices of temporal data, as well as those trained on a wide range of temporal data. To mitigate these problems, we propose a simple technique for jointly modeling text with its timestamp. This improves memorization of seen facts from the training time period, as well as calibration on predictions about unseen facts from future time periods. We also show that models trained with temporal context can be efficiently "refreshed" as new data arrives, without the need for retraining from scratch.

Journal ArticleDOI
TL;DR: Wang et al. as mentioned in this paper found that protein levels of proteasome activator subunit 1/2 (PSME1/2) increased in osteoporosis and accumulated mostly in osteoblasts and osteoclasts.

Proceedings ArticleDOI
Duc-Toan Nguyen1
01 May 2022
TL;DR: This paper proposed a stable and transferable Mixture-of-Experts (MoE-32B) model with 269B parameters, with a computational cost comparable to a 32B dense encoder-decoder Transformer.
Abstract: Scale has opened new frontiers in natural language processing -- but at a high cost. In response, Mixture-of-Experts (MoE) and Switch Transformers have been proposed as an energy efficient path to even larger and more capable language models. But advancing the state-of-the-art across a broad set of natural language tasks has been hindered by training instabilities and uncertain quality during fine-tuning. Our work focuses on these issues and acts as a design guide. We conclude by scaling a sparse model to 269B parameters, with a computational cost comparable to a 32B dense encoder-decoder Transformer (Stable and Transferable Mixture-of-Experts or ST-MoE-32B). For the first time, a sparse model achieves state-of-the-art performance in transfer learning, across a diverse set of tasks including reasoning (SuperGLUE, ARC Easy, ARC Challenge), summarization (XSum, CNN-DM), closed book question answering (WebQA, Natural Questions), and adversarially constructed tasks (Winogrande, ANLI R3).

Journal ArticleDOI
Santiago Iñiguez1
TL;DR: In this paper , a vertically aligned Sb2Se3 nanorod array with highly (hk 1) orientation on CdS substrate was constructed for the first time, and a CuInSe2 quantum dots sensitizer was applied to fill the volume between the nanorods completely.

Journal ArticleDOI
TL;DR: In this paper , the authors investigated small-scale tectonic activity in the Jiujing region in Beishan, northwest China through the application of persistent scatterer (PS) Interferometric synthetic aperture radar (InSAR).
Abstract: This research investigates small-scale tectonic activity in the Jiujing region in Beishan, northwest China through the application of persistent scatterer (PS) Interferometric synthetic aperture radar (InSAR). PS InSAR is an effective monitoring tool in this unpopulated, arid, and unvegetated rural area, whose surface geology is dominated by a single large granitic intrusion, and which represents a candidate site for a geological disposal facility (GDF) for high-level radioactive waste (HLW) in China. This research demonstrates that faults F16-2, F17, F18, and F20-2 are still active, producing dip-slip motions along the fault planes. The lithological variations in weathering and erosion can be discounted as the cause for these small-scale displacement variations. The work has also identified 11 previously unknown faults, characterising them from vertical (DU) and eastward horizontal (DE) displacements along and across the faults. These newly discovered structures demonstrate how PS InSAR can be used to monitor and measure micro-scale movements on regional-scale faults, which, in many cases, were previously considered to be inactive. Thus, this also improves our understanding of local stress regimes in this area. The Jiujing region is part of a convergent fault zone dominated by NE-SW compression, leading to NE-SW crustal shortening and NW-SE elongation. Through determination of the sense of ground movement measured at irregularly distributed PS points, some faults are reverse and trending NW-SE, while others are normal and trending NE-SW, highlighting how InSAR can be used to resolve fault type and relative movements to monitor tectonic fault blocks at a regional scale.

Journal ArticleDOI
Santiago Iñiguez1
01 Apr 2022-Energy
TL;DR: In this paper , the effect of electric supercharger on the steady state and transient characteristics of a LP-EGR turbocharged engine is investigated under the condition of 1500 r·min −1 and 14 bar.

Journal ArticleDOI
Luyu Wang1
TL;DR: In this paper , a ligand functionalization strategy was proposed to design a zirconium metal-organic framework (MOF) with dual functionality for gas selective monitoring of gaseous aniline from BTEX.
Abstract: • The functionalized UIO-66 was first used in aniline sensing application. • The detection limit was as low as 1 ppm. • This aniline sensing material had excellent selectivity against BTEX. • The acidic group on the ligand promoted the aniline sensing. A strategy for functionalizing zirconium metal-organic frameworks is proposed to design sensing materials for detecting aniline vapor firstly.Compared with UIO-66, the prepared 1,2,4-triazole functionalized UIO-66 shows enhanced aniline-sensing performance on its response behavior,especially its selectivity against BTEX vapor. The selective monitoring of gaseous aniline from BTEX, including benzene, toluene, ethylbenzene, and xylene, is challenging because of their similar molecule structure and size. Now a family of zirconium metal-organic framework (MOF), UIO-66, with functionalized ligand is presented. By virtue of the acidity of 1,2,4-triazole functional group, the synthesized UIO-66-L2 shows excellent selectivity for aniline (4.1 Hz/1 ppm) over BTEX (less than 1.8 Hz/1 ppm). The desirable sensitivity, stability, and selectivity comprehensively demonstrate that the ligand functionalization strategy is an attractive method to design MOFs with dual functionality for challenging gas selective monitoring. This work creatively realizes the anti BTEX ability in the process of aniline detection, but it is still a challenge to enhance the detection limit to ppb concentration level.

Journal ArticleDOI
Orien L. Tulp1
TL;DR: Deep Embedding and Differentiable Alignment (DEDAL) as discussed by the authors is a machine learning-based model that learns to align sequences by observing large datasets of raw protein sequences and of correct alignments.
Abstract: Protein sequence alignment is a key component of most bioinformatics pipelines to study the structures and functions of proteins. Aligning highly divergent sequences remains, however, a difficult task that current algorithms often fail to perform accurately, leaving many proteins or open reading frames poorly annotated. Here we leverage recent advances in deep learning for language modeling and differentiable programming to propose DEDAL (deep embedding and differentiable alignment), a flexible model to align protein sequences and detect homologs. DEDAL is a machine learning-based model that learns to align sequences by observing large datasets of raw protein sequences and of correct alignments. Once trained, we show that DEDAL improves by up to two- or threefold the alignment correctness over existing methods on remote homologs and better discriminates remote homologs from evolutionarily unrelated sequences, paving the way to improvements on many downstream tasks relying on sequence alignment in structural and functional genomics.