30 research outputs found
Do Preprocessing and Class Imbalance Matter to the Deep Image Classifiers for COVID-19 Detection? An Explainable Analysis
No full textIn a world withstanding the waves of a raging pandemic, respiratory disease detection from chest radiological images using machine-learning approaches has never been more important for a widely accessible and prompt initial diagnosis. A standard machine-learning disease detection workflow that takes an image as input and provides a diagnosis in return usually consists of four key components, namely input preprocessor, data irregularities (like class imbalance, missing and absent features, etc.) handler, classifier, and a decision explainer for better clarity. In this study, we investigate the impact of the three primary components of the disease-detection workflow leaving only the deep image classifier. We specifically aim to validate if the deep classifiers may significantly benefit from additional preprocessing and efficient handling of data irregularities in a disease-diagnosis workflow. To elaborate, we explore the applicability of seven traditional and deep preprocessing techniques along with four class imbalance handling approaches for a deep classifier, such as ResNet-50, in the task of respiratory disease detection from chest radiological images. While deep classifiers are more capable than their traditional counterparts, explaining their decision process is a significant challenge. Therefore, we also employ three gradient visualization algorithms to explain the decision of a deep classifier to understand how well each of them can highlight the key visual features of the different respiratory diseases
Reply on the comments on the paper “evidence of fluvial to marine transition in the siwalik rocks of the itanagar area, arunachal pradesh, india: Implication for the regional paleogeography” by mullick & sinha 2024, himalayan geology, 45(1), 138-154
No full textWe welcome the comments by Chakraborty et al. on our paper and acknowledge them for their great time and effort for reading our article so thoroughly and to provide their precious suggestions. Here, we reply to the doubts and queries raised by them for clarification, most of which possibly arise due to misinterpretation of our data. We hereby respond to the queries about the technical issues, facies, drainage system and sedimentological interpretation and in reply would definitely like to answer some of their logical queries in light of sedimentological overview and our field observations, keeping in mind not to substantiate the credibility of Indian sedimentologists in front of international researchers. We would like to address the comments in a pointwise manner to make the writing more straight-forward and to-the-point
Study on Integrative Clustering of Multiple Genomic Data to Discover Cancer Subtypes.
No full textWith the advancement of technology, different sources of genetic information become available with a low cost. In the research for finding cancer subtypes, what will help to proceed with a targeted treatment, this opened up a new dimension. However, the basic problem is how to reach towards a proper integration scheme such that both the personal significance and interactive information is conserved, because only then it will be possible to utilize the data resource and obtain richer information about subtypes. On the other hand, as the subtypes are not always properly defined or even known, thus any solution should be unsupervised in nature. This study presented an integration scheme based on the concept of iCluster method, to address these issues. With its many merits, however the crisp nature of clusters obtained by iCluster is not always natural in the case of overlapping and incomplete nature of the data, thus a rough-fuzzy clustering approach will be more suitable, where an addition of intelligent initial center selection algorithm is most desired. A variety of cluster validation index are used to support the claims and present the findings on two different cancer data
Evidence of fluvial to marine transition in the Siwalik rocks of the Itanagar area, Arunachal Pradesh, India: Implication for the regional paleogeography
No full textSedimentological study of the Middle Siwalik Subansiri and lower part of the Upper Siwalik, Siji Formation, from the Itanagar area evidence for marine incursions within an established continental depositional setting. The braided fluvial deposit of the Subansiri Formation is overlain by the shallow marine fan or braid delta deposit of the Siji Formation, recording for the first time the fluvial to marine transition from this area. In the Subansiri Formation isolated large cross strata of the bar platform, downcurrent accreting cross sets of bar of supra-platform, bar top low-stage channel scour fills and rare, thinly rippled silty flood plain deposits were recognized. These assemblages constitute 5-12 m thick sheet sandstone bodies stacked up, forming about a km-thick Subansiri sandstone succession. In the overlying Siji Formation, abundant wave-and combined flow-ripples, well-developed hummocky and swaley strata, and brackish water trace fossils (Lingulichnus, Arenicolites) indicate a marine depositional regime. The depositional domains of the Siji Formation include mudstone dominated prodelta, alternating planar and cross bedded mudstone-sandstone units of the lower delta front; channelized, cross stratified pebbly sandstone and conglomerate of delta mouth bars, and fine sandstone units with tidal bundles from shallow coastal embayment. The Shillong Plateau is a unique feature in the eastern Himalaya whose deformation pattern influenced the depositional system and makes it different from the Western Himalaya. Further a comparison is carried out of the distinctive features of the western Himalayan Siwalik with those of the eastern (Itanagar) region and emphasize that two distinct tectono-geomorphic regimes characterized the foreland basin system
Caste, class and family structure in West Bengal villages
No full textAnalysis of household survey data collected from three villages of West Bengal, India, shows that caste status continues to be significantly related to structure. However, class status - whether based on occupation or landownership - has a stronger & statistically more significant relationship with family structure. Further analysis shows that both occupational classes & caste structure are strongly related to landownership & also show statistically significant relationships with each other. It appears that it is because of their strong relationship with landownership that occupational classes & caste structure maintain significant relationships with family structure. 7 Tables, 22 References. Adapted from the source document.CD: JCFSAOSource type: Electronic(1
Nuclear and joint family households in West Bengal villages
No full textExamines whether nuclear family households are gaining ground in India at the expense of joint family households, based on a questionnaire survey of all 544 households of 3 villages in the Arambagh region of the Hooghly District of West Bengal. The findings suggest that: (1) although the joint family household is not the typical household in rural West Bengal, either in terms of its proportion among all households or in terms of people belonging to it, it shows a remarkable stability, & thus provides support to the cyclical view that joint family households are relatively stable in their rate of incidence, & nuclear family households are transitional forms in the normal functioning of the developmental cycle of joint family units; (2) the developmental cycle of formation & dissolution of joint family households varies by different groups in a village in that the proportion of joint family households tends to be significantly higher among household heads who are age 45+, own land, operate large farms, follow agricultural occupations, belong to high or middle castes, &/or are literate, than among those who are age 44 or younger, landless, operate smaller farms, follow nonagricultural occupations, belong to low castes, or are illiterate; (3) only four characteristics - age, landownership, occupation, & literacy - are independently related to the incidence of household types, & they show a stronger relationship to household types when combined into an index than does each characteristic, except age, related singly; & (4) once the household head is age 45+, regardless of his other characteristics, & his first son is old enough to bring his wife into the household, his probability of heading a joint family increases significantly. The high proportion of nuclear family households in the 3 villages suggest that married sons tend to separate from their parental households almost immediately after they themselves become parents. The joint family household, however, persists as a unit for a relatively longer period of time after it comes into existence among those people in a village who own land, follow agricultural occupations, & are literate. 7 Tables, 16 References. Modified AA.CD: ETNLB6 RX: 1 (on Apr 03, 2007)Source type: Electronic(1
Adaptive Learning-Based -Nearest Neighbor Classifiers With Resilience to Class Imbalance
No full textThe classification accuracy of a k-nearest neighbor (k NN) classifier is largely dependent on the choice of the number of nearest neighbors denoted by k. However, given a data set, it is a tedious task to optimize the performance of k NN by tuning k. Moreover, the performance of k NN degrades in the presence of class imbalance, a situation characterized by disparate representation from different classes. We aim to address both the issues in this paper and propose a variant of k NN called the Adaptive k NN (Ada-k NN). The Ada-k NN classifier uses the density and distribution of the neighborhood of a test point and learns a suitable point-specific k for it with the help of artificial neural networks. We further improve our proposal by replacing the neural network with a heuristic learning method guided by an indicator of the local density of a test point and using information about its neighboring training points. The proposed heuristic learning algorithm preserves the simplicity of k NN without incurring serious computational burden. We call this method Ada-k NN2. Ada-k NN and Ada-k NN2 perform very competitive when compared with k NN, five of k NN\u27s state-of-the-art variants, and other popular classifiers. Furthermore, we propose a class-based global weighting scheme (Global Imbalance Handling Scheme or GIHS) to compensate for the effect of class imbalance. We perform extensive experiments on a wide variety of data sets to establish the improvement shown by Ada-k NN and Ada-k NN2 using the proposed GIHS, when compared with k NN, and its 12 variants specifically tailored for imbalanced classification
Generalized mean based back-propagation of errors for ambiguity resolution
No full textAmbiguity in a dataset, characterized by data points having multiple target labels, may occur in many supervised learning applications. Such ambiguity originates naturally or from misinterpretation, faulty encoding, and/or incompleteness of data. However, most applications demand that a data point be assigned a single label. In such cases, the supervised learner must resolve the ambiguity. To effectively perform ambiguity resolution, we propose a new variant of the popular Multi-Layer Perceptron model, called the Generalized Mean Multi-Layer Perceptron (GMMLP). In GMMLP, a novel differentiable error function guides the back-propagation algorithm towards the minimum distant target for each data point. We evaluate the performance of the proposed algorithm against three alternative ambiguity resolvers on 20 new artificial datasets containing ambiguous data points. To further test for scalability and comparison with multi-label classifiers, 18 real datasets are also used to evaluate the new approach
On Supervised Class-Imbalanced Learning: An Updated Perspective and Some Key Challenges
No full textThe problem of class imbalance has always been considered as a significant challenge to traditional machine learning and the emerging deep learning research communities. A classification problem can be considered as class imbalanced if the training set does not contain an equal number of labeled examples from all the classes. A classifier trained on such an imbalanced training set is likely to favor those classes containing a larger number of training examples than the others. Unfortunately, the classes that contain a small number of labelled instances usually correspond to rare and significant events. Thus, poor classification accuracy on these classes may lead to severe consequences. In this article, we aim to provide a comprehensive summary of the rich pool of research works attempting to combat the adversarial effects of class imbalance efficiently. Specifically, following a formal definition of the problem of class imbalance, we explore the plethora of traditional machine learning approaches aiming to mitigate its adversarial effects. We further discuss the state-of-the-art deep-learning-based approaches for improving a classifier\u27s resilience against class imbalance and highlight the need for techniques tailored for such a paradigm. Moreover, we look at the emerging applications where class imbalance can be a major concern. Finally, we outline a few open problems along with the various challenges emerging with the advent of modern applications, deep learning paradigm, and new sources of data
