ISI Digital Commons (Indian Statistical Institute )
Not a member yet
7571 research outputs found
Sort by
The role of extrinsic and intrinsic factors in perceptual filling-in of the blind-spot with variegated color and texture stimuli
No full textVision scientists dedicated their efforts to unraveling the mechanism of filling-in at the blind-spot (BS) through numerous psychophysical experiments. The prevalent interpretation, emphasizing active filling-in, has spurred extensive research endeavors. In a parallel vein, a pertinent study highlighted the predominance of the nasal Visual Field (VF) over the temporal one and postulated the role of the Cortical Magnification Factor (CMF) in explaining the asymmetry of filling-in. In this study, we first replicated this experiment and then conducted BS-specific psychophysical experiments employing various bi-colored and bi-textured (patterned) stimuli. We observed that nasal dominance is not persistent in the context of the spread of perception for BS filling-in. We posit that the visual information processing priority index (VIPPI), comprising the CMF (an intrinsic factor unaffected by stimulus characteristics) and relative luminance (an extrinsic factor dependent on stimulus characteristics), governs the spread of perception for filling-in in case of diverse neighborhoods of the BS
Transfer Learning in Weather Prediction: Why, How, and What Should
No full textTransfer learning (TL) is a popular phrase in deep learning (DL) domain. It is one of the latest artificial intelligence (AI) technologies that has a significant impact on big data analysis. Methods of traditional machine learning (ML) require the availability of an adequate quantity of training data as well as similarity of characteristics among the feature spaces corresponding to training and test data while performing supervised learning tasks. However, in real-life analytical problems, data scarcity often arises. In such scenarios, the TL approach has shown effectiveness in transferring knowledge from the source tasks that had large training data to a target task that has less training data. Basically, in TL, a model that has been trained on one task is essentially applied to a second related (but not exact) task. In this way, the issue of distribution mismatch can also be addressed. TL is not like conventional machine learning algorithms that try to learn each task starting from the beginning. Meteorological research is such an example of big data analysis which often faces the data scarcity issue. The current study addresses the contemporary challenges in weather forecasting that can be solved (or better dealt with) using TL methods. It presents a brief review of earlier research with the evolution of various technologies used since 1990s, followed by potential applications of TL algorithms to several key challenges in weather prediction, which includes the prediction of air quality, thunderstorms, precipitation, visibility, and cyclones, among others. Special emphasis is given to high-impact weather (HIW) prediction. These high-impact events are extremely difficult to predict, and they can cause enormous property damage and fatalities around the world. TL techniques have shown advantages in predicting HIW. Various challenging issues in implementing TL technology are then discussed. Finally, we address various prospects associated with TL, propose new research directions, and more importantly mention some concerns for beginners in DL-TL research. An extensive list of references is also provided
Unboundedness of the first Betti number and the last Betti number of numerical semigroups generated by concatenation
No full textWe show that the minimal number of generators and the Cohen-Macaulay type of a family of numerical semigroups generated by concatenation of arithmetic sequences is unbounded
Universal penalized regression (Elastic-net) model with differentially methylated promoters for oral cancer prediction
No full textBACKGROUND: DNA methylation showed notable potential to act as a diagnostic marker in many cancers. Many studies proposed DNA methylation biomarker in OSCC detection, while most of these studies are limited to specific cohorts or geographical location. However, the generalizability of DNA methylation as a diagnostic marker in oral cancer across different geographical locations is yet to be investigated. METHODS: We used genome-wide methylation data from 384 oral cavity cancer and normal tissues from TCGA HNSCC and eastern India. The common differentially methylated CpGs in these two cohorts were used to develop an Elastic-net model that can be used for the diagnosis of OSCC. The model was validated using 812 HNSCC and normal samples from different anatomical sites of oral cavity from seven countries. Droplet Digital PCR of methyl-sensitive restriction enzyme digested DNA (ddMSRE) was used for quantification of methylation and validation of the model with 22 OSCC and 22 contralateral normal samples. Additionally, pyrosequencing was used to validate the model using 46 OSCC and 25 adjacent normal and 21 contralateral normal tissue samples. RESULTS: With ddMSRE, our model showed 91% sensitivity, 100% specificity, and 95% accuracy in classifying OSCC from the contralateral normal tissues. Validation of the model with pyrosequencing also showed 96% sensitivity, 91% specificity, and 93% accuracy for classifying the OSCC from contralateral normal samples, while in case of adjacent normal samples we found similar sensitivity but with 20% specificity, suggesting the presence of early disease methylation signature at the adjacent normal samples. Methylation array data of HNSCC and normal tissues from different geographical locations and different anatomical sites showed comparable sensitivity, specificity, and accuracy in detecting oral cavity cancer with across. Similar results were also observed for different stages of oral cavity cancer. CONCLUSIONS: Our model identified crucial genomic regions affected by DNA methylation in OSCC and showed similar accuracy in detecting oral cancer across different geographical locations. The high specificity of this model in classifying contralateral normal samples from the oral cancer compared to the adjacent normal samples suggested applicability of the model in early detection
What calibrating variable-value population ethics suggests
No full textVariable-Value axiologies avoid Parfit\u27s Repugnant Conclusion while satisfying some weak instances of the Mere Addition principle. We apply calibration methods to two leading members of the family of Variable-Value views conditional upon: first, a very weak instance of Mere Addition and, second, some plausible empirical assumptions about the size and welfare of the intertemporal world population. We find that such facts calibrate these two Variable-Value views to be nearly totalist, and therefore imply conclusions that should seem repugnant to anyone who opposes Total Utilitarianism only due to the Repugnant Conclusion
Distributed Graph Computations via Mobile Robots
No full textThe realm of distributed computing by mobile robots on graphs has witnessed significant advancements in recent years, offering solutions to various real-world problems. This research delves into two fundamental problems in this domain, focusing on the development of efficient algorithms and crash tolerance. To date, we have been able to provide (i) efficient, crash-tolerant dispersion algorithms for mobile robots [1], and (ii) the exploration of small dominating sets and maximal independent sets through mobile robots [2]
Minimum Consistent Subset in Trees and Interval Graphs
No full textIn the Minimum Consistent Subset (MCS) problem, we are presented with a connected simple undirected graph G, consisting of a vertex set V (G) of size n and an edge set E(G). Each vertex in V (G) is assigned a color from the set {1, 2, ⋯, c}. The objective is to determine a subset V′ ⊆ V (G) with minimum possible cardinality, such that for every vertex v ∈ V (G), at least one of its nearest neighbors in V′ (measured in terms of the hop distance) shares the same color as v. The decision problem, indicating whether there exists a subset V′ of cardinality at most l for some positive integer l, is known to be NP-complete even for planar graphs. In this paper, we establish that the MCS problem is NP-complete on trees. We also provide a fixed-parameter tractable (FPT) algorithm for MCS on trees parameterized by the number of colors (c) running in O(26cn6) time, significantly improving the currently best-known algorithm whose running time is O(24cn2c+3). In an effort to comprehensively understand the computational complexity of the MCS problem across different graph classes, we extend our investigation to interval graphs. We show that it remains NP-complete for interval graphs, thus enriching graph classes where MCS remains intractable
Mo2E: Mixture of Two Experts for Class-Imbalanced Learning from Medical Images
No full textClass imbalance in the medical image dataset is almost inherent due to the limited availability of clinical data for certain diseases and patient populations. Under-represented classes in the training set affect the classification task because the classifier tends to learn more from the majority classes, which are more common in the dataset and ignore data from the minority classes. To mitigate this issue, we propose a method to learn using two different convolutional neural network-based experts; such experts try to learn boundaries within the head classes, between the head and tail classes, and within the tail classes. During expert training, we integrate the MixUp regularization method to augment imbalanced data, employing distinct data sampling strategies for more effective mixing compared to random selection in traditional MixUp. During the inference phase, we combine the logits of the different experts based on their expertise in the corresponding classes. This way, we can improve the accuracy of the head and tail classes. Experiments using highly imbalanced and long-tailed datasets demonstrate the effectiveness of the suggested framework
Testing Self-Reducible Samplers
No full textSamplers are the backbone of the implementations of any randomised algorithm. Unfortunately, obtaining an efficient algorithm to test the correctness of samplers is very hard to find. Recently, in a series of works, testers like Barbarik, Teq, Flash for testing of some particular kinds of samplers, like CNF-samplers and Horn-samplers, were obtained. But their techniques have a significant limitation because one can not expect to use their methods to test for other samplers, such as perfect matching samplers or samplers for sampling linear extensions in posets. In this paper, we present a new testing algorithm that works for such samplers and can estimate the distance of a new sampler from a known sampler (say, uniform sampler). Testing the identity of distributions is the heart of testing the correctness of samplers. This paper\u27s main technical contribution is developing a new distance estimation algorithm for distributions over high-dimensional cubes using the recently proposed sub-cube conditioning sampling model. Given subcube conditioning access to an unknown distribution P, and a known distribution Q defined over {0, 1}n, our algorithm CubeProbeEst estimates the variation distance between P and Q within additive error ζ using O(n2/ζ4) subcube conditional samples from P. Following the testing-via-learning paradigm, we also get a tester which distinguishes between the cases when P and Q are ε-close or η-far in variation distance with probability at least 0.99 using O(n2/(η − ε)4) subcube conditional samples. The estimation algorithm in the sub-cube conditioning sampling model helps us to design the first tester for self-reducible samplers. The correctness of the testers is formally proved. On the other hand, we implement our algorithm to create CubeProbeEst and use it to test the quality of three samplers for sampling linear extensions in posets
A Unified Deep Learning Framework for Sentiment Analysis of Reviews
No full textOnline user-generated content is increasing rapidly on a daily basis, along with the expansion of social media and e-commerce activities. This contains users’ personal views and opinions in regard to various products, activities, news, ideas, politics, etc., in textual form. Automated analysis of these opinions’ tone helps to make decisions and devise strategies. This activity is known as sentiment analysis or opinion mining and is essential for text processing. We propose a unified framework as a sequential end-to-end process of efficient sentiment analysis. The framework uses a combination of deep learning models to better and efficiently assess the text based on spatial and temporal context and linguistic interrelationships. The proposed framework first improves word embeddings to represent sentiments better. These improved embeddings are used to assess the usefulness of reviews. The identified useful and relevant reviews are retained in the dataset, and the unhelpful reviews are discarded to reduce the dataset size. The reduced dataset is then analyzed for subjectivity sentence-wise, using the improved word embeddings for feature representation. The identified objective sentences are filtered from the reviews, leaving only the subjective sentences. Finally, polarity classification is performed on these obtained reviews to identify their overall sentiment as positive or negative. This framework is tested on two datasets of different domains. The superiority of the results is demonstrated by comparing them to the state-of-the-art sentiment analysis techniques. The framework achieves better performance and outperforms the existing methods