De novo peptide sequencing of tandem mass spec-trometry (MS/MS) has emerged as an important technology for peptide sequencing in proteomics. To increase the accuracy and practicality of de novo sequencing, some previous algorithms used multiple spectra to identify the peptide sequence. In this approach, we focus on de novo sequencing of multiple SILAC-based tandem mass spectra. SILAC technology uses medium containing different kinds of isotope-labeled essential amino acids, usually Arginine(R) and Lysine(K), to label newly synthesized proteins with stable isotopes during cell growth. Multiple MS/MS spectra for the same peptide sequence are produced by spectrometry after the SILAC samples are processed by LC-MS/MS shotgun proteomics. Based on the factors such as the type of isotope labeling, precursor ion mass, etc., multiple spectra with different type of SILAC PTMs for the same peptide can be used to identify the peptide sequence. In this paper, we present two de novo sequencing algorithms to compute the peptide sequence which are based on total number of SILAC modifications and based on the numbers of SILAC Arginine(R) and Lysine(K).
The application of millimeter-wave(mmWave) radar in human activity recognition has attracted significant attention because of its insensitivity to ambient lighting and privacy concerns. Millimeter-wave radar will be transformed into a point cloud as an input. This paper proposes a system that dynamically adjusts DBSCAN to handle point cloud noise. It consists of four main components: random forest to determine minPts, KNN to compute Eps, Cluster Merging, and Noise Re-judgment to optimize the point cloud clustering problem in human activity recognition. The proposed new method makes the dynamic selection of the two parameters minPts and Eps of DBSCAN, as well as the design of a new clustering method for human point cloud clustering based on the property of local sparsity of point cloud of the human body taken by radar. We build a dataset based on a millimeter-wave radar of 10 volunteers. Based on this dataset using the Rand Index and Purity assessment, our proposed method has higher accuracy than other methods, reaching average accuracy of 88.52% and 84.86%, respectively, almost 18% higher than other methods.
The capability to infer emotional insights from emojis found in social media has projected emoji analysis into the spotlight of current emoji-based research. Previous studies mainly used text-surrounding emojis to estimate sentimentality scores. However, trying to conclude the same score based solely on emojis is challenging. In this paper this challenge was welcomed, and with it we created a new concept. This revolutionary scoring method, named the EmojiSets Sentiment Score Rank, proposes using sets of emojis taken from tweets along with information from previous studies [1] to find a sentiment score. This bottom-up scoring approach gives each emoji a sentiment score. It then calculates the context-level sentiment score of a tweet solely dependent on the emojis found within it. To the best of the authors' knowledge, no such approach has been researched in the Emojis Sentiment Analysis area. We tested our model against over 1.2 million tweets concerning Covid-19 and compared it to the VADER model [7] to validate our assumption. Our model corrected around 72% of the tweets that the other model scored as neutral. To succor these findings, 32 human annotators were given the task of annotating 8040 randomly chosen tweets. When calculating similarity using the Jaccard Index, their results were consistent with our approach in over 70% of cases
Thanks to creative rendering and display techniques, 360-degree images can provide a more immersive and interactive experience for streaming users. However, such features make the perceptual characteristics of 360-degree images more complex than those of fixed-view images, making it impossible to achieve a comprehensive and accurate image quality assessment (IQA) task using a simple stack of pre-processing, post-processing, compression, and rendering tasks. In order to thoroughly learn global and local features in 360-degree images, reduce the complexity of multichannel neural network models and simplify the training process, this paper proposes a user-aware joint architecture and an efficient converter dedicated to 360-degree no-reference (NR) IQA. The input of the proposed method is a 360-degree cubic mapping projection (CMP) image. In addition, the proposed 360-degree NRIQA method includes a non-overlapping self-attentive selection module based on a dominant map and a feature extraction module based on a U-shaped transformer (U-former) to address perceptual region significance and projection distortion. The transformer-based architecture and the weighted averaging technique are jointly used to predict local perceptual quality. Experimental results obtained on widely used databases show that the proposed model outperforms other state-of-the-art methods in the case of NR 360-degree image quality assessment. In addition, cross-database evaluation and ablation studies demonstrate the intrinsic robustness and generalization of the proposed model.
It is widely believed that progress towards advanced artificial intelligence (AI) systems will only occur when brain-inspired information processing systems are available to guide autonomous behaviour. The development of such autonomous systems (AS) requires cognitive entities capable of acquiring information, learning, planning, and reasoning that eventually adapt to environmental uncertainties. Cognitive dynamic systems (CDS) provide an engineering tool to design AS. Although the literature addresses various aspects of cognitive dynamic systems, it needs a holistic methodology to analyze alternative techniques concerning the driving module that internally guides the system toward a goal. Since uncertain events far from expectation play a significant role in triggering information-seeking behaviours, contextual surprise can be considered an intrinsic motivator in CDS. This presentation highlights connections and similarities among several definitions of contextual surprise. For demonstration purposes, it reviews the design of a linear Gaussian CDS as a motivating example. It uses a contextual surprise minimization scheme to express information utility and guide CDS’ state estimation and control.
Basic research in Cognitive Informatics (CI) and Cognitive Computing (CC) provides fundamental theories of intelligence science for Autonomous AI (AAI) and cognitive systems. The field of CI and CC has led to general AI technologies triggered by the transdisciplinary advances in brain, intelligence, computer, knowledge, cognitive, robotic, and cybernetic sciences for engineering implementations. This paper presents a summary report of the plenary panel (Part II) on “Recent Advances in Cognitive Informatics and Cognitive Computing towards Autonomous Systems” in the 21 st IEEE International CI/CC Conference (ICCI*CC'22). Strategic CI/CC applications are presented in cognitive systems, AAI, cognitive robots, intelligent vehicles, AI knowledge learning, autonomous intelligence generation, cognitive digital twins, and safety-and-mission-critical systems.
Deep learning and neural networks have been used in various machine learning applications in the past decades. Developing a precise understanding of the underling behavior of neural networks is crucial for their efficient deployment. In this paper, we use an information theoretic approach to study the flow of information in a discrete Hopfield neural network. While discrete Hopfield networks, are used as associative memories, study of their dynamic behavior is important in general, as they represent the family of recurrent neural networks. We determine lower and upper bounds for the entropy and the conditional entropy. We also study the mutual information between the input patterns and the final output of the network. Experimental results support the theoretical conclusions of the paper.
Modern multi-function radar is designed to perform a few functions such as guidance, fire control, communications, and surveillance. It needs schedule many tasks with different properties such as start time, dwell time, priority etc. In this type of radar, the radar resource management module makes decisions in task selection and task scheduling, which are NP-hard problems. Many task scheduling algorithms have been proposed, however it is still very challenging to choose the appropriate algorithm in varying environments. In this work, a modified Q-learning (QL) method, is developed to choose the optimal solution. The modified Q-learning (MQL) method is a reinforcement learning using the paradigm of Deep Q-Network. The MQL considers 8 states and 4 actions (scheduling algorithms) which are used to make decisions. The MQL agent is trained and tested for various episode limits ranging from 500 to 300,000. In each episode, tasks are generated randomly, for the training purpose. A cost function is formulated to compare scheduling performance. Our simulation results show that the proposed approach can choose the best algorithm consistently.
Social media has become a source of information for many people because of its freedom of use. As a result, fake news spread quickly and easily, regardless of its credibility, especially over the past decade. The vast amount of information being shared has fraudulent practices that negatively affect readers' cognitive abilities and mental health. In this study, we aim to introduce a new Arabic COVID-19 dataset for fake news related to COVID-19 from Twitter and Facebook. Afterward, we applied two pre-trained models of classification AraBERT and BERT base Arabic. As a result, AraBERT models obtained better accuracy than BERT base Arabic in two datasets.
To address the current problem of calculating stone grain size in the field of sand and gravel aggreg es, image segmentation of stone targets is achieved by stone images, and the grain length of stone targets is finally obtained. By pre-processing the target stone images, the pre-processed stone images are segmented and predicted using deep learning image processing techniques, and the predicted result maps are subjected to morphological and image binarization operations for subsequent stone particle size calculation. The algorithm is implemented to delineate the assignment of individual stone regions and to find the boundary coordinate points of individual stone image regions, and to calculate the image grain size length of stones from them. The true grain length of the stone is calculated by the proportional mapping relationship between the camera and the pixel length of the stone taken and the real stone length. Through experiments, this operation procedure can segment and calculate the grain length of stones quickly and accurately.
It is recognized that the kernels of human knowledge and theories of abstract sciences have been archived in mathematical forms with indispensable generality and abstractedness [1]–[8]. Intelligent Mathematics (IM) [8]–[14] is contemporary mathematics for manipulating complex mathematical entities in the brain and intelligence systems known as hyperstructures ($\mathbb{H}$) for formalizing rigorous expressions, inferences, and computational intelligence generation. $\mathbb{H}$ [8] are nonnumerical abstract entities beyond those of real numbers ($\mathbb{R}$) and bits ($\mathbb{B}$). Therefore, IM becomes highly demanded across modern Abstract Sciences [15]–[17] as counterparts of classic entity-oriented Concrete Sciences.
Spatially regularized correlation filters (SRCF) have recently received increasing interest for Unmanned Aerial Vehicle (UAV) tracking due to their promising results. While the choice of spatial weight matrices is vital for the success of SRCF methods, they are generally learned only with the training samples in current frame, which results in time-discontinuous spatial weight matrices in neighboring frames, thus degrading the CF models. In this paper, we propose a Smooth Target-Aware Spatially Regularized Correlation Filter (STASRCF) framework for UAV tracking. Specifically, we first obtain the initial target-aware spatial weight matrix in each frame by employing the image segmentation techniques for separating the target from the background, then multiple adaptive spatial regularization terms are integrated into the CF framework for jointly updating the spatial weight matrices and CF models. In this way, time-continuous spatial weight matrices and robust CFs can be learned during tracking, thereby benefiting the tracking performance. In addition, we suggest an Alternating Direction Method of Multipliers (ADMM) method for solving STASRCF efficiently, in which each sub-problem has a closed-form solution. Experiments on multiple UAV datasets show that STASRCF can not only surpass the baseline CSR-DCF by an average AUC gain of 1.9%, but also perform favorably against other state-of-the-art CF trackers.
When given two human face images to choose, a subject's decision process is recorded as a sequence of eye movement gaze points. This sequence is then analyzed to detect and predict the preference made by the subject. In an experiment with twelve subjects, each with sixty trials, we have analyzed the 720 sequences using five attributes and combinatorial fusion. Results are promising with good accuracy and efficiency. In this paper, we characterize the decision-making behavior of each subject and measure the cognitive diversity between each of these twelve subjects and a prototypical subject. Our study contributes to improving the data and predictive quality of the experiment and the computational modeling.
Dealing with the large volumes of data generated in the digital world requires efficient data interpretation techniques. Some compressive feature detection techniques are proposed to efficiently minimize the data size. Most of these algorithms in the signal processing area are based on monoscale or multi-scale methods. These methods are not sensitive to the reshuffling of the data, and critical features like correlation and covariance are lost. An alternative approach is to use polyscale algorithms. This paper provides a polyscale approach using the variance fractal dimension to be developed for compressive detection. Some critical aspects of using these measures are discussed, including robustness to noise and simplicity concerning other existing methods. These measures are applied to some speech utterances and experimental data measured from an electro hydrostatic actuator test rig to show the efficiency of this method. This work can provide insight into practical compressive detections to explore this promising area for future developments.
Optical character recognition (OCR) has many ap-plications, such as digitizing historical documents, automating processes, and helping visually impaired people read. However, extracting text from images into a digital format is not an easy problem to solve, and the outputs from the OCR frameworks often include errors. The complexity comes from the many variations in (digital) fonts, handwriting, lighting, etc. To tackle this problem, this thesis investigates two different methods for correcting the errors in OCR output. The used dataset consists of Swedish addresses. The methods are therefore applied to postal automation to investigate the usage of these methods for further automating postal work by automatically reading addresses on parcels using OCR. The first method, the lexical implementation, uses a dataset of Swedish addresses so that any valid address should be in this dataset (hence there is a known and limited vocabulary), and misspelled addresses are corrected to the address in the lexicon with the smallest Levenshtein distance. The second approach uses the same dataset, but with artificial errors, or artificial noise, added. The addresses with this artificial noise are then used together with their correct spelling to train a machine learning model based on Neural machine translation (NMT) to automatically correct errors in OCR read addresses. The results from this study could contribute by defining in what direction future work connected to OCR and postal addresses should go. The results were that the lexical implementation outperformed the NMT model. However, more experiments including real data would be required to draw definitive conclusions as to how the methods would work in real-life applications.
Tunisia has entered a phase freedom of speech with access to social media since the Jasmine Revolution in 2011. Toxic contents such as abusive and hateful speeches have become omnipresent on Tunisian social media. Considering the side effects of these toxic contents on the psychology of users, it is necessary to detect them automatically. The dialect of Tunisian is underrepresented. As a consequence, there is not enough data set. In this paper, we present the data collection process with the aim of having a Tunisian reference dataset, to evaluate different models of hate speech and abuse detection. We also present our neural network model based on AraBERT. Our experimental results on our dataset shows that the AraBERT model performs better with an F1 score of 0.99.
Autonomous software requirement analysis and specifications is not only an ultimate goal of cognitive computing, but also a persistent challenge to theories and technologies of software engineering. A cognitive computing model is demanded to autonomously elicit and rigorously refine software requirements in order to generate a set of formal specifications. This paper presents a novel methodology for the design of a cognitive computing method for Software Requirement Elicitation and Specifications (SRES) based on the latest advances in software science and intelligent mathematics. SRES is implemented as an interactive system for capturing software requirements and generating formal specifications. The SRES methodology and experiments are demonstrated for solving real-world and complex software engineering problems enabled by cognitive computing theories underpinned by intelligent mathematics.
Basic research in Cognitive Informatics (CI) and Cognitive Computing (CC) provides fundamental theories of intelligence science for Autonomous AI (AAI) and cognitive systems. The field of CI and CC has led to general AI technologies triggered by the transdisciplinary advances in brain, intelligence, computer, knowledge, cognitive, robotic, and cybernetic sciences for engineering implementations. This paper presents a summary of the plenary panel (Part I) on the “Recent Breakthroughs in Cognitive Informatics and Cognitive Computing towards AAI” in the 21st IEEE International ICCI*CC Conference (ICCI*CC'22). Strategic CI/CC applications are presented in cognitive systems, AAI, cognitive robots, intelligent vehicles, AI knowledge learning, autonomous intelligence generation, and safety-and-mission-critical systems.
We present a new regularized autoencoder for robust feature learning. The regularization, implying stochastic sensitivity, is defined as the sum of entries of the absolute covariance matrix of the output perturbation at each layer of the autoencoder. The advantages of the stochastic sensitivity regularization are two-fold. Firstly, we show that the classical Frobenius norm regularization effectively enforces the network to be insensitive to input perturbation and that the Frobenius norm regularization is a special case of the proposed stochastic sensitivity regularization which enables the proposed method to train an autoencoder for robust feature learning. Secondly, we also show that the stochastic sensitivity regularization attempts to drive the network to learn a set of decorrelated feature maps which removes redundant information and thus improves generalization capabilities. These two properties enable the autoencoder to learn a set of robust and diverse feature maps. Finally, the efficacy and the robustness of the proposed regularization method are confirmed a nd quantified by comparing it against existing regularized auto encoders over a range of tasks.
Neuromorphic (brain-inspired) cameras communicate transients in luminance as events, supporting high frame rate, fine temporal resolution, high dynamic range, no motion blur, high signal-to-noise ratio, and power-efficient vision systems. Here, we used a hybrid camera, neuromorphically enhancing frame-based vision with event-driven motion-sensitive adaptive attention. We used spatial-memory and event-driven salient maps to direct a frame camera toward attentive zones. We further implemented a dynamic tolerance mechanism with which attentional zones reallocate, supporting a multiple-object attentional load. Finally, we enhanced our hybrid imaging system using an automatic segmentation performed in fixed time intervals on event-based reconstructed frames, allowing efficient derivation of attentive areas. This hybrid approach demonstrates an efficient coverage of a large field of view.
