Evaluation of Deep Neural Network Compression Methods for Edge Devices Using Weighted Score-Based Ranking Scheme

The demand for object detection capability in edge computing systems has surged. As such, the need for lightweight Convolutional Neural Network (CNN)-based object detection models has become a focal point. Current models are large in memory and deployment in edge devices is demanding. This shows that the models need to be optimized for the hardware without performance degradation. There exist several model compression methods; however, determining the most efficient method is of major concern. Our goal was to rank the performance of these methods using our application as a case study. We aimed to develop a real-time vehicle tracking system for cargo ships. To address this, we developed a weighted score-based ranking scheme that utilizes the model performance metrics. We demonstrated the effectiveness of this method by applying it on the baseline, compressed, and micro-CNN models trained on our dataset. The result showed that quantization is the most efficient compression method for the application, having the highest rank, with an average weighted score of 9.00, followed by binarization, having an average weighted score of 8.07. Our proposed method is extendable and can be used as a framework for the selection of suitable model compression methods for edge devices in different applications.

Contextual Image Illustrator: A Review

Any story or any other literary content is best understood and advertised with the help of pictures. Images are used to arouse reader’s interest and comprehension in the content. The contextual image illustrator will take any content description and will output the ranked images related to that content. The text can be any blog, newspaper article, any story or any other content. The image retrieval process that has been used for this purpose is Text based Image Retrieval, i.e., TBIR. Semantic keywords are extricated from the story; images are looked through an annotated database. Thereafter, an image ranking scheme will determine the relevance of each image. Then the user can choose among the images displayed. A score along with each image will also be displayed representing its relevance to the query. The keywords stemming and stop word removal has been explained in the document. Also, the algorithm that has been designed to determine the score and hence the image’s significance has been calculated. Testing consisting of both unit testing and module testing of the project are explained. Keywords: Keyword Extraction, Image Search, Stemming, Stop word Removal, URL Score, URL Ranking

Dietary Risk Assessment and Ranking of Multipesticides in Dendrobium officinale

The presence of pesticide residues in Dendrobium officinale (D. officinale), a commonly used herbal medicine, has attracted much attention in recent years. Therefore, this study presents the levels of 141 pesticide residues in forty D. officinale samples, which were measured by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS). And we used a deterministic estimate model to assess chronic and acute dietary exposure risk, as well as the cumulative risks for adults, children, and specific groups of consumers. Furthermore, the residual risk of individual pesticides was sorted by adapting the matrix-ranking scheme. In 92.5% of the samples, 43 pesticides were detected, of which difenoconazole had the highest detection frequency. Multiple residues were detected in 85.0% of the samples, and one sample contained even up to 17 pesticides. The chronic hazard quotient (HQc) and the acute hazard quotient (HQa) were far below 100%, and both cumulative chronic and acute hazard indices (HI) did not exceed 100%. The risk scoring scheme showed that four pesticides were considered to pose a comparatively potential high risk, including difenoconazole, carbofuran, fipronil, and emamectin benzoate. The results indicated that the occurrence of pesticide residues in D. officinale could not pose a serious health problem to the public.

Euclidean Distance Based Similarity Measurement and Ensuing Ranking Scheme for Document Search from Outsourced Cloud Data

In this paper, we propose the Euclidean Distance based Similarity Measurement and Ensuing Ranking (EDSMER) scheme to aid effective document search from outsourced cloud data. It is another attempt to find an alternative to binary based approaches. In this approach, the User or the Data owner needs to filter out the suitable keywords for the document and then the index is prepared. To provide security and privacy, both the data and the index are encrypted and moved to the cloud space. The application of Euclidean Distance based Similarity Measurement and Ensuing Ranking (EDSMER) scheme for document searching takes place after the authorized user requests for the documents through query terms. Initially the authorized user sends a query to Cloud Service Provider to retrieve all the documents which are mapped with the keywords provided by him. The proposed algorithm calculates the distance between the query terms and the index terms. The minimum the distance, the more it is closer towards each other and vice-versa. Our Euclidean Distance based Similarity Measurement and Ensuing Ranking (EDSMER) scheme greatly enhances the system functionality by sending the most relevant documents instead of transmitting all documents back. The experimental validations are performed on RFC and FIRE dataset. Through experimental analysis, we prove that our proposed approach is secure and efficient as well as exhibits better recall and precision rate in the IR system to deal with the document-retrieval process.

Probability of distributed surface rupturing occurrence and displacement regression for dip-slip earthquakes

<p>Probabilistic fault displacement hazard analysis (PFDHA) estimates the probability of occurrence and the expected exceedance of on-fault (principal fault rupturing; PF) and off-fault (dist ributed rupturing; DR) surface displacement during an earthquake. Here we concent rate on off-fault rupturing on dip-slip earthquakes, and present an original probability model for the occurrence of DR and for the expected exceedance of displacement dist ribution based on an approach named “slicing” (an alternative to the “gridding” approach commonly used). The method is developed based on the compilation and reappraisal of surface ruptures from 32 historical crustal dip-slip earthquakes, with magnitudes ranging from M<sub>w</sub> 4.9 to 7.9. A ranking scheme is applied to distinguish PF (rank 1) from simple DR (rank 2) and t riggered faulting (rank 3). Thus modellers can use prediction equations based on or excluding ruptures st rongly related to local st ructural setting depending on the site of concern. In the case of a st ructural setting at a site where large-scale bending (rank 21, 22) and pre-existing faults (rank 1.5, 3) is considered irrelevant, modelling can be performed considering only the unpredictable DR (rank 2). To minimize bias due to the incomplete nature of the database, we int roduce the “slicing” approach, which considers that the probability of having a surface rupture within slices parallel to the PF is homogeneous along the st rike of each slice. “Slicing” probabilities, computed as a function of magnitude of the earthquake and distance from the PF, are then combined with Monte Carlo simulations that model the dependence of the probability of occurrence of rupture and exceedance of displacement with the dimensions and position of the site of interest with respect to the PF. Finally, both probabilities are combined with existing predictive equations of exceedance of displacement on the PF to calculate fault-displacement hazard curves for sites of interest.</p>

A new release of the SURE database of earthquake surface ruptures suited to Fault Displacement Hazard Analysis

<p>Fault displacement hazard assessment is based on empirical relationships derived from data of historical surface rupturing earthquakes. This approach is used for land use planning, sizing of lifelines or major sensitive infrastructures located in the proximity of active faults. These relationships provide the probability of occurrence of surface rupture and predict the amount of displacement, both for the main ruptures (principal) and for distributed ones appearing beyond.</p><p>Following the first version of the global database SURE 1.0 (Baize et al., 2019), we are continuing the effort to compile observations from well-documented historical and recent surface faulting events in order to feed and improve empirical relationships. The new SURE2.0 global database consolidates the previous version SURE 1.0 data, rejecting some poorly constrained cases, reviewing some cases already in, and adding well-documented new ones (e.g. Ridgecrest sequence, USA, 2019). In total, the SURE 2.0 database has 46 earthquakes, including 15 normal fault cases, 16 reverse fault cases and 15 strike-slip cases from 1872 to 2019. The magnitude range is from M4.9 to 7.9, with ruptures from 5 to 300 km long.</p><p>SURE 2.0 provides the geometric location and attribute information of rupture segments in a GIS environment and a spreadsheet reports the amplitude and characteristics of deformation, including data sources and its eventual geometric refinement during analysis. In this new version, we completed an essential task to derive attenuation relationships, by classifying each rupture segment and each slip measurement point, using a ranking scheme based on the pattern and amplitude of the observed rupture traces, and considering the structural context and the long-term geomorphology. This distinguishes the principal rupture (class 1), which is the main surface expression of the source of the earthquake. Typically, in the siting study, this class is assigned to the identified active fault. Class 2 features (distributed ruptures) are characterized by shorter lengths and smaller displacements that appear randomly close and around the main rupture. We introduced the distributed main fracture category (class 1.5), which corresponds to the relatively long minor fractures recognized on cumulative structures secondary to the main fault. Class 3 represents triggered slip evidences on remote active faults, clearly not connected with the earthquake causative fault (sympathetic ruptures).</p><p>As was done with reverse fault cases (Nurminen et al., 2020), this new SURE 2.0 version will be used to derive probabilities associated with the rupture distribution  during any type of earthquake.</p>

DRDOCK: A drug repurposing platform integrating automated docking, simulations and a log-odds-based drug ranking scheme

MotivationDrug repurposing, where drugs originally approved to treat a disease are reused to treat other diseases, has received escalating attention especially in pandemic years. Structure-based drug design, integrating small molecular docking, molecular dynamic (MD) simulations and AI, has demonstrated its evidenced importance in streamlining new drug development as well as drug repurposing. To perform a sophisticated and fully automated drug screening using all the FDA drugs, intricate programming, accurate drug ranking methods and friendly user interface are very much needed.ResultsHere we introduce a new web server, DRDOCK, Drug Repurposing DOcking with Conformation-sampling and pose re-ranKing - refined by MD and statistical models, which integrates small molecular docking and molecular dynamic (MD) simulations for automatic drug screening of 2016 FDA-approved drugs over a user-submitted single-chained target protein. The drugs are ranked by a novel drug-ranking scheme using log-odds (LOD) scores, derived from feature distributions of true binders and decoys. Users can submit a selection of LOD-ranked poses for further MD-based binding affinity evaluation. We demonstrated that our platform can indeed recover one of the substrates for nsp16, a cap ribose 2’-O methyltransferase, and recommends XXX, could be repurposed for the COVID19 treatment. All the sampled docking poses and trajectories can be 3D-viewed and played via our web interface. This platform shall be easy-to-use for general scientists and medicinal researchers to carry out drug repurposing within a couple of days which should add value to our timely responses to, particularly, emergent disease outbreaks.Availability and implementationDRDOCK can be freely accessed from https://dyn.life.nthu.edu.tw/drdock/. (Due to the hardware upgrade, the service is NOT available before 2/15, 2021)