151 research outputs found
Evaluation of the Effect of Tubacin (a specific HDAC-6 inhibitor) on a 3D Neurosphere Model of Parkinson’s Disease
No full text2025Parkinson’s Disease (PD) is the second most common neurodegenerative disease, affecting nearly 8.5 million people worldwide. Its current etiopathogenesis is caused by a multitude of etiopathogenetic mechanisms like oxidative stress, mitochondrial dysfunction, -synuclein misfolding, lewy body formation, and neuroinflammation among others. Current treatments primarily focus on Symptomatic therapy via countering dopamine depletion. Isoform specific inhibition of Histone Deacetylase-6 (HDAC-6) has been a long contested therapeutic strategy for Parkinson’s Disease (PD). Previous in-vitro studies report that HDAC-6 inhibition causes nuclear localization of HDAC-6, proteasomal system dysfunction, increased protein aggregation, and apoptosis. On the contrary , most in-vivo studies show that HDAC-6 inhibition can have a protective effect against toxin-induced PD models via countering neuroinflammation, oxidative stress and microglial activation. This reflects a gap in research translation from traditional monolayer cell culture models to animal studies. This study aims to develop a midbrain Neurosphere model for modeling Parkinson’s disease using neural progenitor stem cells (NPCs), and then evaluating the effect of an isoform-specific Histone Deacetylase-6 (HDAC-6) inhibitor on the NPCs. Spheroid culture was initially generated using 3 methods: Hanging drop technique, Rotary cell culture system (RCCS), and Ultra-low attachment (ULA) microplates. These three methods were compared using various metrics to determine which could best fit the requirement for the project. The ULA microplate method was ultimately used to generate & differentiate the NPC spheroids.Differentiation was confirmed by immunostaining for differentiation markers like Forkhead Box Protein 2(FOXA2), Dopamine Transporter (DAT) & G-protein coupled, Inwardly Rectifying Potassium channel receptor-2 (GIRK2). -synuclein containing preformed fibrils (PFFs) were used as a positive control for disease induction. Next, a standard resazurin assay was performed for determining the cytotoxicity of the selected HDAC-6 inhibitor, Tubacin. The effect of Tubacin was characterized on the expression levels of the differentiation markers on the spheroids, using Immunocytochemistry (ICC). Tubacin did not significantly affect the levels of FOXA2 and GIRK2, but it did show a protective effect on the immunofluorescence of DAT against PFF-induced dopaminergic neuron toxicity. Tubacin treatment also lowered the expression of -synuclein, which was quantified using sandwich ELISA. The findings were visually confirmed via the aggresome assay, which studies ‘aggresomes’ which are visual signs of protein misfolding and aggregation, a key cellular biomarker for PD. Tubacin treatment significantly reduced aggresome levels which upregulated in the patient cell line, as well as attenuated the damage caused by prolonged PFF-treatment. Our findings indicate that isoform specific of HDAC-6 have therapeutic application in Parkinson’s disease
NeuFENet: Neural Finite Element Solutions with Theoretical Bounds for Parametric PDEs
No full textWe consider a mesh-based approach for training a neural network to produce field predictions of solutions to parametric partial differential equations (PDEs). This approach contrasts current approaches for "neural PDE solvers" that employ collocation-based methods to make point-wise predictions of solutions to PDEs. This approach has the advantage of naturally enforcing different boundary conditions as well as ease of invoking well-developed PDE theory -- including analysis of numerical stability and convergence -- to obtain capacity bounds for our proposed neural networks in discretized domains. We explore our mesh-based strategy, called NeuFENet, using a weighted Galerkin loss function based on the Finite Element Method (FEM) on a parametric elliptic PDE. The weighted Galerkin loss (FEM loss) is similar to an energy functional that produces improved solutions, satisfies a priori mesh convergence, and can model Dirichlet and Neumann boundary conditions. We prove theoretically, and illustrate with experiments, convergence results analogous to mesh convergence analysis deployed in finite element solutions to PDEs. These results suggest that a mesh-based neural network approach serves as a promising approach for solving parametric PDEs with theoretical bounds.This is a pre-print of the article Khara, Biswajit, Aditya Balu, Ameya Joshi, Soumik Sarkar, Chinmay Hegde, Adarsh Krishnamurthy, and Baskar Ganapathysubramanian. "NeuFENet: Neural Finite Element Solutions with Theoretical Bounds for Parametric PDEs." arXiv preprint arXiv:2110.01601 (2021). Copyright 2021 The Authors. Posted with permission.
Published as Khara, Biswajit, Aditya Balu, Ameya Joshi, Soumik Sarkar, Chinmay Hegde, Adarsh Krishnamurthy, and Baskar Ganapathysubramanian. "Neufenet: Neural finite element solutions with theoretical bounds for parametric pdes." Engineering with Computers (2024): 1-23. doi: https://doi.org/10.1007/s00366-024-01955-7
Fast Certification of Vision-Language Models Using Incremental Randomized Smoothing
No full textA key benefit of deep vision-language models such as CLIP is that they enable zero-shot open vocabulary classification; the user has the ability to define novel class labels via natural language prompts at inference time. However, while CLIP-based zero-shot classifiers have demonstrated competitive performance across a range of domain shifts, they remain highly vulnerable to adversarial attacks. Therefore, ensuring the robustness of such models is crucial for their reliable deployment in the wild. In this work, we introduce Open Vocabulary Certification (OVC), a fast certification method designed for open-vocabulary models like CLIP via randomized smoothing techniques. Given a base "training" set of prompts and their corresponding certified CLIP classifiers, OVC relies on the observation that a classifier with a novel prompt can be viewed as a perturbed version of nearby classifiers in the base training set. Therefore, OVC can rapidly certify the novel classifier using a variation of incremental randomized smoothing. By using a caching trick, we achieve approximately two orders of magnitude acceleration in the certification process for novel prompts. To achieve further (heuristic) speedups, OVC approximates the embedding space at a given input using a multivariate normal distribution bypassing the need for sampling via forward passes through the vision backbone. We demonstrate the effectiveness of OVC on through experimental evaluation using multiple vision-language backbones on the CIFAR-10 and ImageNet test datasets.This is a preprint from Nirala, A. K., A. Joshi, C. Hegde, and S. Sarkar. "Fast Certification of Vision-Language Models Using Incremental Randomized Smoothing." arXiv e-prints (2023): arXiv-2311.
doi: https://doi.org/10.48550/arXiv.2311.09024. Published as Nirala, Ashutosh, Ameya Joshi, Soumik Sarkar, and Chinmay Hegde. "Fast Certification of Vision-Language Models Using Incremental Randomized Smoothing." In 2024 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML), pp. 252-271. IEEE, 2024.
doi: https://doi.org/10.1109/SaTML59370.2024.00019
Encoding Invariances in Deep Generative Models
No full textReliable training of generative adversarial networks (GANs) typically require massive datasets in order to model complicated distributions. However, in several applications, training samples obey invariances that are \textit{a priori} known; for example, in complex physics simulations, the training data obey universal laws encoded as well-defined mathematical equations. In this paper, we propose a new generative modeling approach, InvNet, that can efficiently model data spaces with known invariances. We devise an adversarial training algorithm to encode them into data distribution. We validate our framework in three experimental settings: generating images with fixed motifs; solving nonlinear partial differential equations (PDEs); and reconstructing two-phase microstructures with desired statistical properties. We complement our experiments with several theoretical results.This is a pre-print of the article Shah, Viraj, Ameya Joshi, Sambuddha Ghosal, Balaji Pokuri, Soumik Sarkar, Baskar Ganapathysubramanian, and Chinmay Hegde. "Encoding Invariances in Deep Generative Models." arXiv preprint arXiv:1906.01626 (2019). Posted with permission.</p
Binary 2D Morphologies of Polymer Phase Separation: Dataset and Python Toolbox
No full textStudy of the intricate connection between the design of material distributions (also called morphology or microstructure) and the final properties of the material system has been an attractive research theme for material science community. Such analysis provides ability to synthesize the microstructures exhibiting desired properties. This theme encompasses several material systems including porous materials [26], steels and welds [2], composites [14], powder metallurgy [28], 3D printing [22], energy storage devices as batteries [10], and energy converting devices like bulk hetero-junction solar cells [20]. Microstructure-sensitive design has been used to tailor a wide variety of properties including strengths, heat and mass diffusivities, energy storage capacity and lifetime, and energy conversion efficiency.This is a manuscript of the article Shah, Viraj, Ameya Joshi, Balaji Sesha Sarath Pokuri, Sambuddha Ghosal, Soumik Sarkar, Baskar Ganapathysubramanian, and Chinmay Hegde. "Binary 2D Morphologies of Polymer Phase Separation: Dataset and Python Toolbox." (2019). Posted with permission.</p
Effect Of Alkali Ferrocyanides On Crystallisation Of Sodium Chloride: Preliminary results
No full textSodium chloride (NaCl) is one of the ubiquitous soluble salts in the environment and is responsible for weathering of building materials. The salt weathering is attributed to the stress developed from crystallisation of these salts in pores of the building materials, with supersaturation as the driving force. In the last years, researchers have successfully mitigated the damage associated with the crystallisa-tion of NaCl by the use of alkali-ferrocyanides (crystallisation inhibitors) in porous building materials. The observed mitigation of the damage has been attributed to lowering of the crystallisation pressure, possibly related to changes in the crystal habit and preferential crystallisation of the salt in the form of efflorescence instead of crypto-florescence. However, the effect of the inhibitor on the development of the so-called crystallisation pressure has not been studied in detail yet. In fact, direct measurement of this pressure is challenging and, until now, only a few experiments have been successful. In this research, an experimental setup has been developed to directly measure the crystallisation forces of NaCl and the effect of fer-rocyanide on these, while visualizing the crystallization process under a microscope. Some preliminary tests using this setup have been carried out: these consisted in monitoring force evolution from a drop of solution with and without the inhibitor confined between two glass plates.Heritage & TechnologyMaterials and Environmen
Differentiable Programming for Piecewise Polynomial Functions
No full textWe introduce a new, principled approach to extend gradient-based optimization to piecewise smooth models, such as k-histograms, splines, and segmentation maps. We derive an accurate form of the weak Jacobian of such functions and show that it exhibits a block-sparse structure that can be computed implicitly and efficiently. We show that using the redesigned Jacobian leads to improved performance in applications such as denoising with piecewise polynomial regression models, datafree generative model training, and image segmentation.This proceeding is published as Cho, Minsu, Ameya Joshi, Xian Yeow Lee, Aditya Balu, Adarsh Krishnamurthy, Baskar Ganapathysubramanian, Soumik Sarkar, and Chinmay Hegde. "Differentiable Programming for Piecewise Polynomial Functions." NeurIPS Thirty-fourth Annual Conference on Neural Information Processing Systems. Learning Meets Combinatorial Algorithms (LMCA): Workshop at NeurIPS 2020. December 6-12, 2020. Posted with permission.</p
Field Solutions of Parametric PDEs
No full textWe consider mesh based approaches for training neural network to produce field predictions to parametric partial differential equations (PDEs). This is in contrast to current approaches for ‘neural PDE solvers’ that employ collocation approaches to make point predictions of PDEs. Our approach has advantages of (a) easier handling of various boundary conditions, and (b) ease of invoking well developed PDE theory – including analysis of numerical stability and convergence – on discretized domains. On the other hand, an obvious disadvantage is the network size required for producing field solutions. We explore such a strategy using two loss functions based on (i) Finite Difference Method (FDM) and (ii) Finite Element Method (FEM) on two canonical parametric PDEs. While the FDM loss is closely related to losses used in recent PINN type approaches, the weighted galerkin loss (FEM loss) is akin to an energy functional that produces improved solutions, satisfies a priori mesh convergence, and can model Neumann boudary conditions. These results suggest that mesh based neural networks are promising approaches for parametric PDEs.This is a pre-print of the article Khara, Biswajit, Aditya Balu, Ameya Joshi, Adarsh Krishnamurthy, Soumik Sarkar, Chinmay Hegde, and Baskar Ganapathysubramanian. "Field Solutions of Parametric PDEs." (2021). Copyright © 2021 The Authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). Posted with permission
DriveCLIP: Zero-shot transfer for distracted driving activity understanding using CLIP
No full textDistracted driving action recognition from naturalistic driving is crucial for both
driver and pedestrian’s safe and reliable experience. However, traditional computer
vision techniques sometimes require a lot of supervision in terms of a large amount
of annotated training data to detect distracted driving activities. Recently, the
vision-language models have offered large-scale visual-textual pre-training that can
be adapted to unsupervised task-specific learning like distracted activity recognition.
The contrastive image-text pretraining models like CLIP have shown significant
promise in learning natural language-guided visual representations. In this paper,
we propose a CLIP-based driver activity recognition framework that predicts
whether a driver is distracted or not while driving. CLIP’s vision embedding offers
zero-shot transfer, which can identify distracted activities by the driver from the
driving videos. Our result suggests this framework offers SOTA performance on
zero-shot transfer for predicting the driver’s state on three public datasets. We also
developed DriveCLIP, a classifier on top of the CLIP’s visual representation for
distracted driving detection tasks, and reported the results here.This is a manuscript of a proceeding published as Hasan, Md Zahid, Ameya Joshi, Mohammed Rahman, Archana Venkatachalapathy, Anuj Sharma, Chinmay Hegde, and Soumik Sarkar. "DriveCLIP: Zero-shot transfer for distracted driving activity understanding using CLIP." Machine Learning for Autonomous Driving Workshop at the 36th Conference on Neural Information Processing
Systems (NeurIPS 2022), New Orleans, USA.
Copyright 2022 The Authors.
Posted with permission
- …
