High-dimensional data in agriculture and remote sensing suffer from the curse of dimensionality, leading to poor model performance and high computational cost. Feature selection mitigates this by removing irrelevant and redundant features while preserving discriminative power. We propose Fuzzy Hypergraph Feature Association Map (FH-FAM), a novel supervised feature selection method that uses fuzzy hypergraphs to model higher-order feature interactions and uncertainty. It computes multi-way normalized mutual information for relevance, multi-way correlation for redundancy, applies sigmoidal and gamma fuzzy membership functions, constructs weighted fuzzy hypergraphs, and selects an optimized subset via maximal independent set after three-stage refinement. FH-FAM was evaluated on 15 public datasets (food/agriculture and remote sensing domains) using Random Forest classification (80:20 train–test split). Compared to FFAMFS, CFS, DSCA, FCFB, FROT, and ABESS, FH-FAM achieved the highest mean accuracy (81.43

Transformer-based language models such as BERT having 110M+ parameters have revolutionized natural language understanding, yet their internal mechanisms remain largely opaque to researchers and practitioners. Traditional attention-based interpretability methods often emphasize structurally important but semantically weak tokens such as punctuation marks rather than meaningful semantic relationships. This work introduces a lightweight and model-agnostic framework for quantifying token-level representational importance using hidden-state activation strengths at Layer 8 of BERT. The proposed Activation Flow Network (AFN) framework computes Token Activation Strength using the L2 norm of Layer-8 hidden representations, enabling direct ranking of semantically salient tokens. The study further introduces a threshold-based activation bucket formulation that partitions tokens into HIGH-activation and LOW-activation groups using an empirical upper-quartile activation boundary. Experimental observations demonstrate that semantically meaningful content words consistently occupy the HIGH-activation bucket and dominate representational activation shifts, while structurally supportive tokens contribute comparatively less. The results suggest that Layer 8 acts as a critical semantic consolidation zone balancing structural and semantic information processing. By revealing how activation magnitudes concentrate around semantically informative tokens, this work provides an interpretable and computationally efficient alternative to attentioncentric analysis, contributing toward transforming BERT from a "black box" into a more transparent "glass box" model for natural language understanding.

Lack of transparency in AI systems poses challenges in critical real-life applications. It is important to be able to explain the decisions of an AI system to ensure trust on the system. Explainable AI (XAI) algorithms play a vital role in achieving this objective. In this paper, we are proposing a new algorithm for Explaining AI systems, FAMeX (Feature Association Map based eXplainability). The proposed algorithm is based on a graph-theoretic formulation of the feature set termed as Feature Association Map (FAM). The foundation of the modelling is based on association between features. The proposed FAMeX algorithm has been found to be better than the competing XAI algorithms - Permutation Feature Importance (PFI) and SHapley Additive exPlanations (SHAP). Experiments conducted with eight benchmark algorithms show that FAMeX is able to gauge feature importance in the context of classification better than the competing algorithms. This definitely shows that FAMeX is a promising algorithm in explaining the predictions from an AI system

The emergence timing of a plant, i.e., the time at which the plant is first visible from the surface of the soil, is an important phenotypic event and is an indicator of the successful establishment and growth of a plant. The paper introduces a novel deep-learning based model called EmergeNet with a customized loss function that adapts to plant growth for coleoptile (a rigid plant tissue that encloses the first leaves of a seedling) emergence timing detection. It can also track its growth from a time-lapse sequence of images with cluttered backgrounds and extreme variations in illumination. EmergeNet is a novel ensemble segmentation model that integrates three different but promising networks, namely, SEResNet, InceptionV3, and VGG19, in the encoder part of its base model, which is the UNet model. EmergeNet can correctly detect the coleoptile at its first emergence when it is tiny and therefore barely visible on the soil surface. The performance of EmergeNet is evaluated using a benchmark dataset called the University of Nebraska-Lincoln Maize Emergence Dataset (UNL-MED). It contains top-view time-lapse images of maize coleoptiles starting before the occurrence of their emergence and continuing until they are about one inch tall. EmergeNet detects the emergence timing with 100% accuracy compared with human-annotated ground-truth. Furthermore, it significantly outperforms UNet by generating very high-quality segmented masks of the coleoptiles in both natural light and dark environmental conditions.

PDF file - 869K, Effect of DNA-PKcs or ATM inhibition on p-T2609 in NSCLCs. h pretreatment with 10 muM NU7441 or 10 muM KU55933. Square boxes represent cropped images shown in Figure 6C

This review focuses on the state-of-the-art algorithms for explainability of artificial intelligence (XAI) systems using different techniques. A plethora of explainability techniques have evolved which are useful for language processing, text classification and other computational text analytics tasks. Various research ideas are to be elevated to improve the potential of the feature extraction techniques in text analytics to endeavor the explanation of the prediction toward more standard, optimum, robust and trustworthy. The survey presented in this article consolidates the related research papers on explainability on text analytics using deep learning techniques and other frameworks.

SummaryOne of the multiple threats to protected areas worldwide, invasive plant species have the potential to decrease biodiversity and ecosystem function. We studied changes in infestation by two widespread invasive plant species – Chromolaena odorata and Mikania micrantha – in India’s Manas National Park, a critical conservation site for threatened flora and fauna. Based on field surveys in 2011 and 2019, we found that C. odorata and M. micrantha were present in most of the sampled plots and had newly invaded over 20% of the plots. However, the abundance of M. micrantha decreased in 45% of the plots while C. odorata increased in >50% of the plots. We used a decision tree-based regression with environmental variables as predictors to generate the distribution, abundance and invasion risk maps of the two species. Among environmental variables, road proximity and fire frequency had the strongest influences, respectively, on C. odorata and M. micrantha. Invaded quadrats exhibited lower native-plant diversity than non-invaded quadrats, and C. odorata specifically had a strong negative association with native-plant community structure. These invasive species have increased their range and abundance, and our predicted invasion risk maps indicate the areas where management intervention is urgently needed.

Crop pests and diseases are major threats to food security globally. The mung bean (Vigna Radiata) is one of the leading crops in India. A large part of the population in India is completely dependent on mung bean . So, high production efficiency for the mung bean is required, which does not happen due to the excessive damage from pests and diseases. Recently, with the advancement of Deep Learning techniques, remarkable performance has been achieved in the field of image classification by employing Convolutional Neural Networks (CNNs). This brings a lot of promise in the field of pest and disease identification by effective image classification. In this paper, we have proposed a novel deep learning-based technique to identify the mung bean pest and disease. In order to handle the limitation arising due to less number of mung bean crop images for the purpose of training, we have adopted transfer learning, which is able to generate a very promising result for quick and easy pest and disease detection. The developed model has successfully recognized 6 different types of mung bean diseases and 4 types of pests out of healthy and affected leaves collected in different seasons. Based on the experiments conducted, the proposed smartphone-based deep learning model for the mung bean pest and disease detection has achieved an average accuracy of 93.65%.

•In recent years, immense work has been done on sentiment analysis directed towards the binary (positive, negative) or ternary (positive, neutral, negative) classification. Due to the complexity of misidentifying hidden sentiments of languages, there is a lack of research in the context of multi-class sentiment classification.•Bengali language is the state language of Bangladesh spoken by 228 million native speakers in 2021. Performing multi-class sentiment classification on Bangla social media comments is difficult owing to its highly inflected characteristics, with lack of punctuation, misspellings and uses of non-standard abbreviations.•Therefore, we performed multi-class sentiment classification on Bengali social media comments to predict into four distinct classes. We used machine learning classifiers such as logistic regression, stochastic gradient descent, decision tree, random forest, multinomial naïve bayes, and support vector machine to obtain higher accuracy than previous research work.•Our dataset consists of 42,036 Facebook comments labeled into four classes: sexual, religious, acceptable, and political. We preprocessed the texts by performing steps like tokenization, stop words removal, stemming, part of speech tagging, and many more.•The results are evaluated using standard evaluation measures like accuracy, precision, recall, and f1 score. To get a deep insight into the model performance, text analysis, confusion matrix, and K-fold cross-validation were also used.

A phenotype is the composite of an observable expression of a genome for traits in a given environment. The trajectories of phenotypes computed from an image sequence and timing of important events in a plant's life cycle can be viewed as temporal phenotypes and indicative of the plant's growth pattern and vigor. In this paper, we introduce a novel method called FlowerPhenoNet, which uses deep neural networks for detecting flowers from multiview image sequences for high-throughput temporal plant phenotyping analysis. Following flower detection, a set of novel flower-based phenotypes are computed, e.g., the day of emergence of the first flower in a plant's life cycle, the total number of flowers present in the plant at a given time, the highest number of flowers bloomed in the plant, growth trajectory of a flower, and the blooming trajectory of a plant. To develop a new algorithm and facilitate performance evaluation based on experimental analysis, a benchmark dataset is indispensable. Thus, we introduce a benchmark dataset called FlowerPheno, which comprises image sequences of three flowering plant species, e.g., sunflower, coleus, and canna, captured by a visible light camera in a high-throughput plant phenotyping platform from multiple view angles. The experimental analyses on the FlowerPheno dataset demonstrate the efficacy of the FlowerPhenoNet.

With size of datasets varying un-uniformly in sample size and feature length, to optimize the feature set usually different methods such as filter, wrapper methods are used. However, with different machine learning techniques though either feature reduction is used, or feature extraction is used, both have its own merits and de-merits. The proposed work proposes a hybrid model that tries to combine the feature extraction and feature reduction techniques thereby using both linear and non-linear techniques to take the best parts of both methods. After the initial ensemble is created still the feature set is further optimized by using the concept of entropy and information gain. Using mutual information, on further analysis the best non-redundant feature sets are selected after considering a specific threshold and using this as a testing tool the datasets are again analyzed to check the working accuracy. The model performance is found to be effective even using reduced feature sub-set. Also, it has been found apart from excelling in classification accuracy, the model has been successful in maintaining the range of the metric irrespective of the input size.

Software vulnerability poses a significant security threat to the simultaneous expansion of the digital revolution. With increasing numbers of software and their vulnerabilities, detecting vulnerabilities accurately is a substantial challenge. Various static and deep learning approaches are executed to make the tasks more manageable, but detection accuracy is still a significant factor. In this paper, we are introducing Common Vulnerability Detector (CVD), a deep learning-based vulnerability detection system that can analyze Object-Oriented Programming (OOP) Language assembled source codes and can detect vulnerabilities with the highest accuracy. We implemented a highly optimized Convolutional Recurrent Neural Network (CRNN) for source code analysis to achieve this. By implementing this model on a SARD dataset of C Sharp source codes, CVD could successfully detect six common and dangerous vulnerabilities with an accuracy of 96.10% and F1 score of 96.40%. We compared CVD with all the known and popular methods and CVD outperformed all of them. According to the performance and results, our proposed CVD model is a promising step in vulnerability detection. Furthermore, this model can be the stair for something revolutionary in the world of vulnerability detection.

Autism Spectrum Disorder (ASD) individuals experience many troubles in communication and lag in social improvement. In this significant circumstance, the appearance of smart technologies has kept promising success regarding Autism Spectrum Disorder (ASD). Many socio-developmental mobile applications have been created in developed countries. In contrast, there are no such applications in developing countries like Bangladesh in native content. In this paper, an Android-based mobile application named HELLO is designed with the fundamental thought of Picture Exchange Communication System (PECS) in the Bangla language. This application aims to improve the socio-communicational learnability of autism individuals in Bangladesh. It is expected that our application can be brilliant support to autism individuals, also included the people who endure non-verbal inconveniences. The outcome of this application demonstrates a positive streamline to the improvement in communicating with others.

Bangla Handwritten digit and character recognition, a complex computer vision problem that is important for the Bengali language as the progress in this segment for the Bengali language is slow. We used two popular datasets, BanglaLekha-Isolated and NumbtaDB, for both digits and characters and used a Convolutional neural network to train our model. We augmented our dataset using a shifting method and ran multiple experiments on vowels, digits, and characters. The result is 96.42% average accuracy on BanglaLekha augmented. Our model also achieved 98.92% accuracy on the NumtaDB dataset. We used our model to sketch up two models, License plate recognition and Smart E-learning application. We used connected component analysis in License plate recognition that helped us to extract essential segments of the license plate. We used Keras as a TensorFlow backend in our research. Bangla OCR research is ongoing and will get better over time with better datasets and learning techniques.

While immunotherapy with checkpoint blockade for advanced non-small cell lung cancer (NSCLC) has become first-line in select patients, durable responses are only seen in 15-20% of cases. As such, there is an urgent need to identify new therapeutic targets to overcome immunosuppression in the tumor microenvironment (TME). Activation of the Wnt/β-Catenin pathway in non-small cell lung cancer (NSCLC) is associated with tumor growth, metastasis, and polarization of tumor-associated macrophages to an immunosuppressive M2-like phenotype.

In today's world, around 230 million people used the Bengali or Bangla language to communicate. These individuals are progressively associated with online exercises on famous micro-blogging and long-range interpersonal communication locales, imparting insights what's more, musings, and also the vast majority of articles are in the Bengali language. Thus, Bengali people express their emotions using the Bangla language by reviewing, commenting, or recommendations. Sentiment analysis helps determine the people's emotions expressed on social media or several online platforms. Therefore, this study focused on extracting their emotion from a Bengali text by utilizing Word2vector, Skip-Gram, and Continuous Bag of Words (CBOW) with a new Word to Index model by focusing on three individual classes happy, angry, and excited. The authors achieved the highest accuracy of 75% by utilizing the skip-gram model to classify those three types of emotions. This study also outperformed other existing works with LSTM, CNN model with existing datasets.

High dimensionality of input dataset is an important issue in machine learning. Though there are a few effective ways of dimensionality reduction, none of them, other than feature selection, preserve the original features of the input dataset. Feature selection assesses relevance of features along with their potential redundancy for dimensionality reduction. This paper proposes a first of its kind fuzzy graph based technique for measuring feature relevance and redundancy. The proposed technique addresses the uncertainty in interpretation of feature relevance and redundancy by defining a fuzzy set based formulation of feature relevance and redundancy. Superior performance of the proposed algorithm is illustrated by simulation experiments with 20 benchmark datasets. The proposed algorithm has demonstrated an average feature reduction of 86% along with a 10% improvement in average classification accuracy compared to the state-of-the-art algorithms.

Improving multilingual end-to-end (E2E) automatic speech recognition (ASR) systems have manifold advantages. They simplify the training strategy, are easier to scale and exhibit better performance over monolingual models. However, it is still challenging to use a single multilingual model to recognize multiple languages without knowing the input language, as most multilingual models assume the availability of the input language. In this paper, we introduce multi-softmax model to improve the multilingual recurrent neural network transducer (RNN-T) models, by having language specific softmax, joint and embedding layers, while sharing rest of the parameters. We extend the multi-softmax model to work without knowing the input language, by integrating a language identification (LID) model, that estimates the LID on-the-fly and also does the recognition at the same time. The multi-softmax model outperforms monolingual models with an average word error rate relative (WERR) reduction of 4:65% on Indian languages. Finetuning further improves the WERR reduction to 12:2%. The multisoftmax model with on-the-fly LID estimation, shows WERR reduction of 13:86% compared to the multilingual baseline.

Bangladesh is a developing country with a massive population. Therefore, government hospitals are always overcrowded with patients. Very few people can afford the cost of private hospitals. Even health workers are also concentrated only in urban hospitals. Hence, people had to face many challenges to get health care services. Recently, there is a revolutionary change in the Information and Communication (ICT) sector in Bangladesh. So, the telemedicine and eHealth service is getting closer to individuals level. In this paper, we have explored the recent status of telemedicine and E-health services in Bangladesh in a pragmatic way. We founded out the communication medium, lackings of health care sites and challenging factors for the people who use this service. Our experimental analysis reveals most of the people still unconcerned about this service. We have found problems like language barrier, inadequate authentication, improper interface, etc. in the telemedicine and E-health sites. By resolving those issues, more people will engage with this service.

To nudge the state of the art of human-machine interacting applications, research in speech recognition systems has progressively been examining speech-to-text synthesis, but implementation has been done to minimal languages. Although the Bengali language has not been much of an object of interest, we present the automatic speech recognition (ASR) system solely based on this particular language since around 16% of the world’s population speak Bengali. It has been a demanding task to implement Bengali ASR because it consists of diacritic characters. We conduct a series of preprocessing and feature selection methods along with a convolutional neural net model in consideration of an automatic verbal communication recognition system. Furthermore, the researchers compared this method to a recurrent neural network that is based on an LSTM network and a vast data file of Google Inc. Investigation of these two models indicates such as the recurrent neural net outperforms the convolutional neural net: the former benefits from combining connectionist temporal classification (CTC) and language model (LM). A quantitative analysis of the output shows that the word error rate and validation loss can be affected by variation in dropout values. It also shows that the parameters are also affected by clean and augmented data.