Abstract

Abstract Background: Automatic functional annotation of proteins is an open research problem in bioinformatics. The growing number of protein entries in public databases, for example in UniProtKB, poses challenges in manual functional annotation. Manual annotation requires expert human curators to search and read related research articles, interpret the results, and assign the annotations to the proteins. Thus, it is a time-con‑ suming and expensive process. Therefore, designing computational tools to perform automatic annotation leveraging the high quality manual annotations that already exist in UniProtKB/SwissProt is an important research problem Results: In this paper, we extend and adapt the GrAPFI (graph-based automatic protein function inference) (Sarker et al. in BMC Bioinform 21, 2020; Sarker et al., in: Pro‑ ceedings of 7th international conference on complex networks and their applications, Cambridge, 2018) method for automatic annotation of proteins with gene ontology (GO) terms renaming it as GrAPFI-GO. The original GrAPFI method uses label propaga‑ tion in a similarity graph where proteins are linked through the domains, families, and superfamilies that they share. Here, we also explore various types of similarity measures based on common neighbors in the graph. Moreover, GO terms are arranged in a hier‑ archical manner according to semantic parent–child relations. Therefore, we propose an efcient pruning and post-processing technique that integrates both semantic similarity and hierarchical relations between the GO terms. We produce experimental results comparing the GrAPFI-GO method with and without considering common neighbors similarity. We also test the performance of GrAPFI-GO and other annotation tools for GO annotation on a benchmark of proteins with and without the proposed pruning and post-processing procedure. Conclusion: Our results show that the proposed semantic hierarchical post-pro‑ cessing potentially improves the performance of GrAPFI-GO and of other annotation tools as well. Thus, GrAPFI-GO exposes an original efcient and reusable procedure, to exploit the semantic relations among the GO terms in order to improve the automatic annotation of protein functions

Abstract

Autonomous driving vehicles (ADVs) are sleeping giant intelligent machines that perceive their environment and make driving decisions. Most existing ADSs are built as hand-engineered perception-planning-control pipelines. However, designing generalized handcrafted rules for autonomous driving in an urban environment is complex. An alternative approach is imitation learning (IL) from human driving demonstrations. However, most previous studies on IL for autonomous driving face several critical challenges: (1) poor generalization ability toward the unseen environment due to distribution shift problems such as changes in driving views and weather conditions; (2) lack of interpretability; and (3) mostly trained to learn the single driving task. To address these challenges, we propose a view-invariant spatiotemporal attentive planning and control network for autonomous vehicles. The proposed method first extracts spatiotemporal representations from images of a front and top driving view sequence through attentive Siamese 3DResNet. Then, the maximum mean discrepancy loss (MMD) is employed to minimize spatiotemporal discrepancies between these driving views and produce an invariant spatiotemporal representation, which reduces domain shift due to view change. Finally, the multitasking learning (MTL) method is employed to jointly train trajectory planning and high-level control tasks based on learned representations and previous motions. Results of extensive experimental evaluations on a large autonomous driving dataset with various weather/lighting conditions verified that the proposed method is effective for feasible motion planning and control in autonomous vehicles.

Abstract

Image Fusion maximizes the visual information at each pixel location by merging content from multiple images in order to produce an enhanced image. Exposure Fusion, specifically, fuses a bracketed exposure stack of poorly lit images to generate a properly illuminated image. Given a single input image, exposure fusion can still be employed on a ‘simulated’ exposure stack, leading to direct single image contrast and low-light enhancement. In this work, we present a novel ‘Quaternion Factorized Simulated Exposure Fusion’ (QFSEF) method by factorizing an input image into multiple illumination consistent layers. To this end, we use an iterative sparse matrix factorization scheme by representing the image as a two-dimensional pure quaternion matrix. Theoretically, our representation is based on the dichromatic reflection model and accounts for the two scene illumination characteristics by factorizing each progressively generated image into separate specular and diffuse components. We empirically prove the advantages of our factorization scheme over other exposure simulation methods by using it for the low-light image enhancement task. Furthermore, we provide three exposure fusion strategies which can be used with our simulated stack and provide a comprehensive performance analysis. Finally, in order to validate our claims, we show extensive qualitative and quantitative comparisons against relevant state-of-the-art solutions on multiple standard datasets along with relevant ablation analysis to support our proposition. Our code and data are publicly available for easy reproducibility

Abstract

In this paper, we address the problem of monocular 3D human reconstruction with an acute focus on the challenge of recovering person-specific facial geometry as well as suppressing surface noise, specifically addressing the issue of false geometrical variations caused by textural edges. Most of the existing state-of-the-art methods in this domain fail to address these challenges. More specifically, we propose to integrate facial and wrinkle map priors in a learning-based framework to improve the quality of full-body 3D reconstruction from monocular images. By incorporating facial prior, we recover person-specific identity unlike many of the existing methods which rely on parametric shape models. Similarly, the wrinkle map prior enables our network to alleviate the challenge of false geometrical variations caused by high-frequency textural details present in the input image. We evaluate our method on publicly available datasets & in-the-wild internet images with loose clothing and report superior performance both qualitatively and quantitatively when compared with SOTA methods

Abstract

Posts on Twitter allow users to express ideas and opinions very dynamically. This high volume of data provides relevant clues about the public judgment on a specific product, event, service, etc. While traditional sentiment analysis primarily focuses on classifying the sentiment in general (positive or negative) or at an aspect level (very positive, low negative, and so on) and cannot exploit the intensity information. Recently, the problem of irony detection in social media has been proven to be pervasive among research enthusiasts, posing a challenge to sentiment analysis systems. Moreover, the figurative use of language has received scarce attention from the computational linguistic research point of view. This paper proposes an architecture, the Experts Model, inspired by the standard Mixture of Experts (MoE) model. The key idea here is that each expert learns different sets of features from the feature vector, which helps in better irony detection (Ironic vs. Non-ironic - SemEval-2018 subtask A) and ironic type detection (Verbal irony with vs. without polarity contrast vs. Situational irony vs. Non-irony - SemEval-2018 subtask B) from the tweet. We compared our Experts Model’s results with baseline results along with the top five performers of SemEval-2018 Task-3, Ironic detection. The experimental results show that our proposed approach deals with the ironic detection problem and stands at the top-3 results.

Abstract

The need for automated fact-checking is getting prominent with every passing day as the spread of misinformation is swelling over the ever-increasing stream of online content. We focus on fine-grained labelling of factual information in tweets to facilitate better fact-checking systems capable of providing improved justifications. In this paper, we present a token-level annotation of factual claims in tweets from Indian Twitter. To deal with the multilingual variety of the Indian diaspora, we deal with tweets in English, Bengali, Hindi, and their codemixed variants. To the best of our knowledge, this dataset is first of kind, both in terms of labelling scheme as well as data sources.

Abstract

Grading student work is critical for assessing their understanding and providing necessary feedback. However, answer grading can become monotonous for teachers. On the standard ASAG data set, our system shows substantial improvements in classification disparity of correct and incorrect answers from a reference answer compared to baseline methods. Our supervised model (1) utilizes recent advances in semantic word embeddings and (2) implements ideas from one-shot learning methods, which are proven to work with minimal. We present experimental results from a model based on different approaches and demonstrates decent performance on standard benchmark dataset.

Abstract

Although the city of Hyderabad in Telangana, India lies in seismic zone II, low to medium intensity tremors that pose a serious concern towards safety of the built envi- ronment are not uncommon. One such series of tremors occurred during 13–20 October 2020, in the financial district of Hyderabad and created a panic situation due to perceivable shaking and jolts with loud sounds associ- ated with hydro-seismicity. To understand the safety of the city’s built environment, a study was conducted on low, medium and tall buildings using ground motions recorded at the International Institute of Information Technology (IIIT), Hyderabad, which is 2.3 km from the epicentre. The amplification of ground motion on the second floor of the Nilgiri Building in IIIT, Hyderabad was 1.2–2.3. The vibrations recorded on the ground floor of the Nilgiri Building were used to develop a site-speci- fic response spectrum. This was further used to obtain the peak responses of the considered buildings through response spectrum analysis. The present study suggests that the low-rise buildings, mid-rise buildings and non- structural elements in high-rise buildings are under threat in the case of high-intensity earthquakes.

Abstract

Domain Generalization (DG) requires a model to learn a hypothesis from multiple distributions that generalizes to an unseen distri- bution. Recent explorations show that, for neural networks, the choice of hyper-parameters and model architecture significantly affects DG performance, and making the right choice is non-trivial. In this paper, we show evidence suggesting that the models that perform better at DG, might be implicitly learning a low dimen- sional representation in the feature space. Furthermore, we take forward this idea and employ explicit feature learning to improve DG. To this end, we propose a DG specific supervised contrastive loss. We show how this performance improvement correlates to reduced dimensionality of the representation. Our work establishes new state-of-the-art on five different DG benchmarks, compared against over two dozen existing approaches in DomainBed. We show how this performance improvement correlates to reduced dimensionality of the representation

Abstract

The Internet of Things (IoT) relies heavily on exchange of data collected by nodes in a sensor network. This network data can be represented as a signal on a graph allowing tools from graph signal processing (GSP) to analyse it. Here, we utilize graph signal reconstruction algorithms to extrapolate missing sensor data from partially sampled network data. An important class of graph signal reconstruction algorithms operate by partitioning the graph vertices into local sets to achieve lower reconstruction errors and faster convergence. However, methods for identification of these local neighbourhoods have not been clearly studied in the literature. Here, we propose two flexible algorithms to generate local sets on a graph. These algorithms are based on the distance between the sampled and unsampled vertices. Our algorithms are competitive with the methods in literature while offering flexibility in the number of sampled vertices. We carry out simulation-based analysis of sampling strategies and proposed local set generation algorithms using local-set-based reconstruction algorithms.