Underwater Fish Detection and Counting Using Mask Regional Convolutional Neural Network

Fish production has become a roadblock to the development of fish farming, and one of the issues encountered throughout the hatching process is the counting procedure. Previous research has mainly depended on the use of non-machine learning-based and machine learning-based counting methods and so was unable to provide precise results. In this work, we used a robotic eye camera to capture shrimp photos on a shrimp farm to train the model. The image data were classified into three categories based on the density of shrimps: low density, medium density, and high density. We used the parameter calibration strategy to discover the appropriate parameters and provided an improved Mask Regional Convolutional Neural Network (Mask R-CNN) model. As a result, the enhanced Mask R-CNN model can reach an accuracy rate of up to 97.48%.

Electronic Voting System Using an Enterprise Blockchain

Conventional electronic voting systems use a centralized scheme. A central administration of these systems manages the entire voting process and has partial or total control over the database and the system itself. This creates some problems, accidental or intentional, such as possible manipulation of the database and double voting. Many of these problems have been solved thanks to permissionless blockchain technologies in new voting systems; however, the classic consensus method of such blockchains requires specific computing power during each voting operation. This has a significant impact on power consumption, compromises the efficiency and increases the system latency. However, using a permissioned blockchain improves efficiency and reduces system energy consumption, mainly due to the elimination of the typical consensus protocols used by public blockchains. The use of smart contracts provides a secure mechanism to guarantee the accuracy of the voting result and make the counting procedure public and protected against fraudulent actions, and contributes to preserving the anonymity of the votes. Its adoption in electronic voting systems can help mitigate part of these problems. Therefore, this paper proposes a system that ensures high reliability by applying enterprise blockchain technology to electronic voting, securing the secret ballot. In addition, a flexible network configuration is presented, discussing how the solution addresses some of the security and reliability issues commonly faced by electronic voting system solutions.

EXPRESS: The Dominance of Item Learning in the Location-Specific Proportion Congruence Paradigm

Prior research has shown that various cues are exploited to reactively adjust attention and such adjustments depend on learning associations between cues and proportion congruence. This raises the intriguing question of what will be learned when more than one cue is available, a question that has implications for understanding which cue(s) will dominate in guiding reactive adjustments. Using a picture-word Stroop task, Bugg, Suh, Colvett, and Lehmann (2020) provided initial evidence that item learning dominated over location learning in a location-specific proportion congruence (LSPC) paradigm, a pattern that may explain the difficulty researchers have faced in replicating and reproducing the LSPC effect. One goal was to reproduce this pattern using a non-overlapping two-item sets design that more closely matched prior studies, and another goal was to examine generalizability of the pattern to two other tasks. Using a prime-probe, color-word Stroop task (Experiment 1) and a flanker task (Experiment 2), we again found clear dominance of item learning. In Experiment 3, we attempted to disrupt item learning and promote location learning by using a counting procedure that directed participants’ attention to location. Once again, we found the same pattern of item dominance. Additionally, in none of the experiments did we find evidence for conjunctive (location-item) learning. Collectively, the findings suggest item learning is neither design- or task-specific; rather, it is robust, reliable, and not easily disrupted. Discussion centers on factors dictating dominance of item- over location-based adjustments and implications for the broader literature on LSPC effects.

Root bundles and towards exact matter spectra of F-theory MSSMs

Motivated by the appearance of fractional powers of line bundles in studies of vector-like spectra in 4d F-theory compactifications, we analyze the structure and origin of these bundles. Fractional powers of line bundles are also known as root bundles and can be thought of as generalizations of spin bundles. We explain how these root bundles are linked to inequivalent F-theory gauge potentials of a G4-flux.While this observation is interesting in its own right, it is particularly valuable for F-theory Standard Model constructions. In aiming for MSSMs, it is desired to argue for the absence of vector-like exotics. We work out the root bundle constraints on all matter curves in the largest class of currently-known F-theory Standard Model constructions without chiral exotics and gauge coupling unification. On each matter curve, we conduct a systematic “bottom”-analysis of all solutions to the root bundle constraints and all spin bundles. Thereby, we derive a lower bound for the number of combinations of root bundles and spin bundles whose cohomologies satisfy the physical demand of absence of vector-like pairs.On a technical level, this systematic study is achieved by a well-known diagrammatic description of root bundles on nodal curves. We extend this description by a counting procedure, which determines the cohomologies of so-called limit root bundles on full blow-ups of nodal curves. By use of deformation theory, these results constrain the vector-like spectra on the smooth matter curves in the actual F-theory geometry.

A Comprehensive Comparison Between Different Multiaxial Cycle Counting Procedure

High-temperature nuclear design codes, such as Section III, Division 5 of the American ASME Boiler and Pressure Vessel Code and the French RCC-MRx, require evaluating fatigue damage for qualifying high-temperature components. Both codes provide clear guidance for counting cycles under uniaxial loading conditions, but neither provides a cycle counting procedure for multiaxial loading conditions. The ASTM E1049 also does not address multiaxial cycle counting. However, several widely utilized multiaxial cycle counting procedures are available in the open literature, but there is no agreement on the most appropriate method for high-temperature applications. Applying the different cycle counting methods to the same loading history generally produces different results. Comparisons between cycle counting procedures are available for low-temperature high-cycle fatigue but not for high-temperature low-cycle dwell-fatigue applications. This work presents an extensive comparison between different multiaxial cycle counting procedures potentially suitable for high-temperature low-cycle dwell-fatigue applications. Furthermore, how to conservatively assemble design transients to construct a loading history is also an open question. This work also investigates the uncertainty related to the loading sequence. The results guide the selection of the most appropriate cycle counting procedure, strain range metric, and cycle distribution for ASME Section III, Division 5 applications.

Are small additions solved by direct retrieval from memory or automated counting procedures? A rejoinder to Chen and Campbell (2018)

Contrary to the longstanding and consensual hypothesis that adults mainly solve small single-digit additions by directly retrieving their answer from long-term memory, it has been recently argued that adults could solve small additions through fast automated counting procedures. In a recent article, Chen and Campbell (Psychonomic Bulletin & Review, 25, 739–753, 2018) reviewed the main empirical evidence on which this alternative hypothesis is based, and concluded that there is no reason to jettison the retrieval hypothesis. In the present paper, we pinpoint the fact that Chen and Campbell reached some of their conclusions by excluding some of the problems that need to be considered for a proper argumentation against the automated counting procedure theory. We also explain why, contrary to Chen and Campbell’s assumption, the network interference model proposed by Campbell (Mathematical Cognition, 1, 121–164, 1995) cannot account for our data. Finally, we clarify a theoretical point of our model.

Learning to count: A deep learning framework for graphlet count estimation

Graphlet counting is a widely explored problem in network analysis and has been successfully applied to a variety of applications in many domains, most notatbly bioinformatics, social science, and infrastructure network studies. Efficiently computing graphlet counts remains challenging due to the combinatorial explosion, where a naive enumeration algorithm needs O(Nk) time for k-node graphlets in a network of size N. Recently, many works introduced carefully designed combinatorial and sampling methods with encouraging results. However, the existing methods ignore the fact that graphlet counts and the graph structural information are correlated. They always consider a graph as a new input and repeat the tedious counting procedure on a regular basis even if it is similar or exactly isomorphic to previously studied graphs. This provides an opportunity to speed up the graphlet count estimation procedure by exploiting this correlation via learning methods. In this paper, we raise a novel graphlet count learning (GCL) problem: given a set of historical graphs with known graphlet counts, how to learn to estimate/predict graphlet count for unseen graphs coming from the same (or similar) underlying distribution. We develop a deep learning framework which contains two convolutional neural network models and a series of data preprocessing techniques to solve the GCL problem. Extensive experiments are conducted on three types of synthetic random graphs and three types of real-world graphs for all 3-, 4-, and 5-node graphlets to demonstrate the accuracy, efficiency, and generalizability of our framework. Compared with state-of-the-art exact/sampling methods, our framework shows great potential, which can offer up to two orders of magnitude speedup on synthetic graphs and achieve on par speed on real-world graphs with competitive accuracy.

Summer Abundance of Harbour Porpoises (Phocoena phocoena) in the Coastal Waters of Iceland and the Faroe Islands

This study presents the first fully corrected abundance estimates for the harbour porpoise (Phocoena phocoena) for Iceland and the Faroe Islands. In both regions reliable estimates are needed to assess the impact of by-catch and other threats to harbour porpoises. Aerial surveys with harbour porpoise as a secondary and main target species were conducted in the summers of 2007 and 2010 in Icelandic and in Faroese coastal waters respectively. In Iceland, the cue counting procedure was used (which also produces the data required for line transect analysis), while in the Faroese, standard line transect sampling was applied, following the SCANS-II (Small Cetacean Abundance in the North Sea) survey protocol. In both surveys, identical aircraft surveyed at an altitude of 600 ft and a speed of 90–100 kn. Only data collected during Beaufort Sea States (BSS) lower than 3 and during good or moderate porpoise sighting conditions were used for abundance estimates. Abundance estimates were corrected using stratified estimates of esw (incorporating g(0)) values derived during the SCANS-II survey in 2005 as principal observers took part in this survey as well. In Iceland, realised effort in good or moderate harbour porpoise sighting conditions totalled 8,289 km in 13 survey strata, where 77 sightings (109 individuals) were made by the experienced harbour porpoise observer only. In Faroese waters, only part of the area inside the 300 m depth curve could be surveyed and 1,564 km were surveyed in good or moderate porpoise sighting conditions, yielding 39 sightings (49 individuals). The total abundance estimates were 43,179 porpoises (CV=0.45; 95% CI: 31,755–161,899) for Icelandic coastal waters and 5,175 porpoises (CV=0.44; 95% CI: 3,457–17,637) for Faroese waters.