Showing 1–14 of 14 results for author: Ambekar, S

GeneralizeFormer: Layer-Adaptive Model Generation across Test-Time Distribution Shifts

Authors: Sameer Ambekar, Zehao Xiao, Xiantong Zhen, Cees G. M. Snoek

Abstract: We consider the problem of test-time domain generalization, where a model is trained on several source domains and adjusted on target domains never seen during training. Different from the common methods that fine-tune the model or adjust the classifier parameters online, we propose to generate multiple layer parameters on the fly during inference by a lightweight meta-learned transformer, which w… ▽ More We consider the problem of test-time domain generalization, where a model is trained on several source domains and adjusted on target domains never seen during training. Different from the common methods that fine-tune the model or adjust the classifier parameters online, we propose to generate multiple layer parameters on the fly during inference by a lightweight meta-learned transformer, which we call \textit{GeneralizeFormer}. The layer-wise parameters are generated per target batch without fine-tuning or online adjustment. By doing so, our method is more effective in dynamic scenarios with multiple target distributions and also avoids forgetting valuable source distribution characteristics. Moreover, by considering layer-wise gradients, the proposed method adapts itself to various distribution shifts. To reduce the computational and time cost, we fix the convolutional parameters while only generating parameters of the Batch Normalization layers and the linear classifier. Experiments on six widely used domain generalization datasets demonstrate the benefits and abilities of the proposed method to efficiently handle various distribution shifts, generalize in dynamic scenarios, and avoid forgetting. △ Less

Submitted 15 February, 2025; originally announced February 2025.

Comments: WACV 2025

FedGAT: A Privacy-Preserving Federated Approximation Algorithm for Graph Attention Networks

Authors: Siddharth Ambekar, Yuhang Yao, Ryan Li, Carlee Joe-Wong

Abstract: Federated training methods have gained popularity for graph learning with applications including friendship graphs of social media sites and customer-merchant interaction graphs of huge online marketplaces. However, privacy regulations often require locally generated data to be stored on local clients. The graph is then naturally partitioned across clients, with no client permitted access to infor… ▽ More Federated training methods have gained popularity for graph learning with applications including friendship graphs of social media sites and customer-merchant interaction graphs of huge online marketplaces. However, privacy regulations often require locally generated data to be stored on local clients. The graph is then naturally partitioned across clients, with no client permitted access to information stored on another. Cross-client edges arise naturally in such cases and present an interesting challenge to federated training methods, as training a graph model at one client requires feature information of nodes on the other end of cross-client edges. Attempting to retain such edges often incurs significant communication overhead, and dropping them altogether reduces model performance. In simpler models such as Graph Convolutional Networks, this can be fixed by communicating a limited amount of feature information across clients before training, but GATs (Graph Attention Networks) require additional information that cannot be pre-communicated, as it changes from training round to round. We introduce the Federated Graph Attention Network (FedGAT) algorithm for semi-supervised node classification, which approximates the behavior of GATs with provable bounds on the approximation error. FedGAT requires only one pre-training communication round, significantly reducing the communication overhead for federated GAT training. We then analyze the error in the approximation and examine the communication overhead and computational complexity of the algorithm. Experiments show that FedGAT achieves nearly the same accuracy as a GAT model in a centralised setting, and its performance is robust to the number of clients as well as data distribution. △ Less

Submitted 20 December, 2024; originally announced December 2024.

Selective Test-Time Adaptation for Unsupervised Anomaly Detection using Neural Implicit Representations

Authors: Sameer Ambekar, Julia A. Schnabel, Cosmin I. Bercea

Abstract: Deep learning models in medical imaging often encounter challenges when adapting to new clinical settings unseen during training. Test-time adaptation offers a promising approach to optimize models for these unseen domains, yet its application in anomaly detection (AD) remains largely unexplored. AD aims to efficiently identify deviations from normative distributions; however, full adaptation, inc… ▽ More Deep learning models in medical imaging often encounter challenges when adapting to new clinical settings unseen during training. Test-time adaptation offers a promising approach to optimize models for these unseen domains, yet its application in anomaly detection (AD) remains largely unexplored. AD aims to efficiently identify deviations from normative distributions; however, full adaptation, including pathological shifts, may inadvertently learn the anomalies it intends to detect. We introduce a novel concept of selective test-time adaptation that utilizes the inherent characteristics of deep pre-trained features to adapt selectively in a zero-shot manner to any test image from an unseen domain. This approach employs a model-agnostic, lightweight multi-layer perceptron for neural implicit representations, enabling the adaptation of outputs from any reconstruction-based AD method without altering the source-trained model. Rigorous validation in brain AD demonstrated that our strategy substantially enhances detection accuracy for multiple conditions and different target distributions. Specifically, our method improves the detection rates by up to 78% for enlarged ventricles and 24% for edemas. △ Less

Submitted 25 October, 2024; v1 submitted 4 October, 2024; originally announced October 2024.

Comments: MICCAIw ADSMI 2024 [Best Paper Award]

Topological engineered 3D printing of Architecturally Interlocked Petal-Schwarzites

Authors: Rushikesh S. Ambekar, Leonardo V. Bastos, Douglas S. Galvao, Chandra S. Tiwary, Cristiano F. Woellner

Abstract: The topologically engineered complex Schwarzites architecture has been used to build novel and unique structural components with a high specific strength. The mechanical properties of these building blocks can be further tuned, reinforcing with stronger and high surface area architecture. In the current work, we have built six different Schwarzites structures with multiple interlocked layers, whic… ▽ More The topologically engineered complex Schwarzites architecture has been used to build novel and unique structural components with a high specific strength. The mechanical properties of these building blocks can be further tuned, reinforcing with stronger and high surface area architecture. In the current work, we have built six different Schwarzites structures with multiple interlocked layers, which we named architecturally interlocked petal-schwarzites (AIPS). These complex structures are 3D printed into macroscopic dimensions and compressed using uniaxial compression. The experimental results show a strong dependency of mechanical response on the number of layers and topology of the layers. Fully atomistic molecular dynamics compressive simulations were also carried out, and the results are in good agreement with experimental observations. They can explain the underlying AIPS mechanism of high specific strength and energy absorption. The proposed approach opens a new perspective on developing new 3D-printed materials with tunable and enhanced mechanical properties. △ Less

Submitted 10 July, 2023; originally announced July 2023.

Comments: 22 pages and 9 figures

Probabilistic Test-Time Generalization by Variational Neighbor-Labeling

Authors: Sameer Ambekar, Zehao Xiao, Jiayi Shen, Xiantong Zhen, Cees G. M. Snoek

Abstract: This paper strives for domain generalization, where models are trained exclusively on source domains before being deployed on unseen target domains. We follow the strict separation of source training and target testing, but exploit the value of the unlabeled target data itself during inference. We make three contributions. First, we propose probabilistic pseudo-labeling of target samples to genera… ▽ More This paper strives for domain generalization, where models are trained exclusively on source domains before being deployed on unseen target domains. We follow the strict separation of source training and target testing, but exploit the value of the unlabeled target data itself during inference. We make three contributions. First, we propose probabilistic pseudo-labeling of target samples to generalize the source-trained model to the target domain at test time. We formulate the generalization at test time as a variational inference problem, by modeling pseudo labels as distributions, to consider the uncertainty during generalization and alleviate the misleading signal of inaccurate pseudo labels. Second, we learn variational neighbor labels that incorporate the information of neighboring target samples to generate more robust pseudo labels. Third, to learn the ability to incorporate more representative target information and generate more precise and robust variational neighbor labels, we introduce a meta-generalization stage during training to simulate the generalization procedure. Experiments on seven widely-used datasets demonstrate the benefits, abilities, and effectiveness of our proposal. △ Less

Submitted 1 July, 2024; v1 submitted 8 July, 2023; originally announced July 2023.

Comments: Accepted by CoLLAs 2024

Mechanical Energy Absorption of Architecturally Interlocked Petal-Schwarzites

Authors: Leonardo V. Bastos, Rushikesh S. Ambekar, Chandra S. Tiwary, Douglas S. Galvao, Cristiano F. Woellner

Abstract: We carried out fully atomistic reactive molecular dynamics simulations to study the mechanical behavior of six newly proposed hybrid schwarzite-based structures (interlocked petal-schwarzites). Schwarzites are carbon crystalline nanostructures with negative Gaussian curvature created by mapping a TPMS (Triply Periodic Minimal Surface) with carbon rings containing six to eight atoms. Our simulation… ▽ More We carried out fully atomistic reactive molecular dynamics simulations to study the mechanical behavior of six newly proposed hybrid schwarzite-based structures (interlocked petal-schwarzites). Schwarzites are carbon crystalline nanostructures with negative Gaussian curvature created by mapping a TPMS (Triply Periodic Minimal Surface) with carbon rings containing six to eight atoms. Our simulations have shown that petal-schwarzite structures can withstand uni-axial compressive stress up to the order of GPa and can be compressed past 50 percent strain without structural collapse. Our most resistant hierarchical structure has a calculated compressive strength of 260~GPa and specific energy absorption (SEA) of 45.95 MJ/kg, while possessing a mass density of only 685 kg/m$^3$. These results show that these structures could be excellent lightweight materials for applications that require mechanical energy absorption. △ Less

Submitted 5 July, 2023; originally announced July 2023.

SKDCGN: Source-free Knowledge Distillation of Counterfactual Generative Networks using cGANs

Authors: Sameer Ambekar, Matteo Tafuro, Ankit Ankit, Diego van der Mast, Mark Alence, Christos Athanasiadis

Abstract: With the usage of appropriate inductive biases, Counterfactual Generative Networks (CGNs) can generate novel images from random combinations of shape, texture, and background manifolds. These images can be utilized to train an invariant classifier, avoiding the wide spread problem of deep architectures learning spurious correlations rather than meaningful ones. As a consequence, out-of-domain robu… ▽ More With the usage of appropriate inductive biases, Counterfactual Generative Networks (CGNs) can generate novel images from random combinations of shape, texture, and background manifolds. These images can be utilized to train an invariant classifier, avoiding the wide spread problem of deep architectures learning spurious correlations rather than meaningful ones. As a consequence, out-of-domain robustness is improved. However, the CGN architecture comprises multiple over parameterized networks, namely BigGAN and U2-Net. Training these networks requires appropriate background knowledge and extensive computation. Since one does not always have access to the precise training details, nor do they always possess the necessary knowledge of counterfactuals, our work addresses the following question: Can we use the knowledge embedded in pre-trained CGNs to train a lower-capacity model, assuming black-box access (i.e., only access to the pretrained CGN model) to the components of the architecture? In this direction, we propose a novel work named SKDCGN that attempts knowledge transfer using Knowledge Distillation (KD). In our proposed architecture, each independent mechanism (shape, texture, background) is represented by a student 'TinyGAN' that learns from the pretrained teacher 'BigGAN'. We demonstrate the efficacy of the proposed method using state-of-the-art datasets such as ImageNet, and MNIST by using KD and appropriate loss functions. Moreover, as an additional contribution, our paper conducts a thorough study on the composition mechanism of the CGNs, to gain a better understanding of how each mechanism influences the classification accuracy of an invariant classifier. Code available at: https://github.com/ambekarsameer96/SKDCGN △ Less

Submitted 1 September, 2022; v1 submitted 8 August, 2022; originally announced August 2022.

Comments: Accepted at ECCV 2022 Workshop VIPriors

Topologically engineered 3D printed architectures with superior mechanical strength

Authors: R. S. Ambekar, B. Kushwaha, P. Sharma, F. Bosia, M. Fraldi, N. Pugno, C. S. Tiwary

Abstract: Materials that are lightweight yet exhibit superior mechanical properties are of compelling importance for several technological applications that range from aircrafts to household appliances. Lightweight materials allow energy saving and reduce the amount of resources required for manufacturing. Researchers have expended significant efforts in the quest for such materials, which require new conce… ▽ More Materials that are lightweight yet exhibit superior mechanical properties are of compelling importance for several technological applications that range from aircrafts to household appliances. Lightweight materials allow energy saving and reduce the amount of resources required for manufacturing. Researchers have expended significant efforts in the quest for such materials, which require new concepts in both tailoring material microstructure as well as structural design. Architectured materials, which take advantage of new engineering paradigms, have recently emerged as an exciting avenue to create bespoke combinations of desired macroscopic material responses. In some instances, rather unique structures have emerged from advanced geometrical concepts (e.g. gyroids, menger cubes, or origami/kirigami-based structures), while in others innovation has emerged from mimicking nature in bio-inspired materials (e.g. honeycomb structures, nacre, fish scales etc.). Beyond design, additive manufacturing has enabled the facile fabrication of complex geometrical and bio-inspired architectures, using computer aided design models. The combination of simulations and experiments on these structures has led to an enhancement of mechanical properties, including strength, stiffness and toughness. In this review, we provide a perspective on topologically engineered architectured materials that exhibit optimal mechanical behaviour and can be readily printed using additive manufacturing. △ Less

Submitted 1 June, 2022; originally announced June 2022.

Journal ref: MATERIALS TODAY (2021), 48, 72-94

Rain energy harvesting using atomically thin Gadolinium Telluride decorated 3D Printed nanogenerator

Authors: Partha Kumbhakar, Arko Parui, Rushikesh S. Ambekar, Madhubanti Mukherjee, Saif Siddique, Nicola M. Pugno, Abhisek K. Singh, Chandra S. Tiwary

Abstract: The 3D printing technology offers an innovative approach for developing energy storage devices to create facile and low-cost customized electrodes for modern electronics. Generating electric potential by moving a droplet of ionic solution over two-dimensional (2D) materials is a novel method for rain energy harvesting. This work demonstrated a liquid-solid contact electrification-based 3D printed… ▽ More The 3D printing technology offers an innovative approach for developing energy storage devices to create facile and low-cost customized electrodes for modern electronics. Generating electric potential by moving a droplet of ionic solution over two-dimensional (2D) materials is a novel method for rain energy harvesting. This work demonstrated a liquid-solid contact electrification-based 3D printed nanogenerator where raindrop passes through the positively charged ultrathin Gadolinium Telluride (Gd2Te3) sheets. Experimental results showed that voltage as high as ~0.6 V could be generated by moving a droplet of ionic solution on the decorated 3D printed nanogenerator. The output efficiency of the nanogenerator is increased ~400% by enhancing the surface area of copious 3D printed porous structures. Density Functional Theory (DFT) calculations are done, revealing that the high electrical conductivity of (112) surface of Gd2Te3 is due to the p-type charge carriers. Additionally, we illustrate the enhancement of the output performance (~0.8V) by using a graphite rod and arbitrarily manipulating the surface charge. Therefore, this work can open up a new avenue to advance scientific research of Blue energy harvesting and tackle the energy crisis. △ Less

Submitted 18 January, 2022; originally announced February 2022.

Schwarzite and schwarzynes based load-bear resistant radial cellular griding-based 3D printed structures

Authors: Eliezer F. Oliveira, Rushikesh S. Ambekar, Douglas S. Galvao, Chandra S. Tiwary

Abstract: Nature-occurring structures exhibiting unique topological features such as complex and gradient porosity has been the basis to create new materials and/or structures. Most studies have been focused on complex periodic porous structures but gradient porous ones have not been yet fully investigated for stable structural designs. In this work, we have proposed and tested a new approach to create cell… ▽ More Nature-occurring structures exhibiting unique topological features such as complex and gradient porosity has been the basis to create new materials and/or structures. Most studies have been focused on complex periodic porous structures but gradient porous ones have not been yet fully investigated for stable structural designs. In this work, we have proposed and tested a new approach to create cellular griding structures, in which the mass density varies from the center to the borders, i.e, a radial gradient. To create these new structures we exploited the topology of two carbon-based families with different pore sizes, the schwarzites, and schwarzynes. We created fully atomistic models that were translated into macroscale ones that were then 3D printed. The mechanical behavior of the gradient structures was investigated by molecular dynamics simulations and mechanical compression tests of the printed models. Our results show that their mechanical response can be engineered (for instance, in terms of energy absorption, ballistic performance, etc.) and can outperform their corresponding density uniform structures. △ Less

Submitted 12 November, 2021; originally announced November 2021.

Comments: 26 pages, 7 figures, and 2 tables

Mechanical Properties of 3D-Printed Pentadiamond

Authors: Levi C. Felix, Rushikesh S. Ambekar, Cristiano F. Woellner, Brijesh Kushwaha, Varinder Pal, Douglas S. Galvao, Chandra S. Tiwary

Abstract: In this work, We combined fully atomistic molecular dynamics and finite elements simulations with mechanical testings to investigate the mechanical behavior of atomic and 3D-printed models of pentadiamond. Pentadiamond is a recently proposed new carbon allotrope, which is composed of a covalent network of pentagonal rings. Our results showed that the stress-strain behavior is almost scale-independ… ▽ More In this work, We combined fully atomistic molecular dynamics and finite elements simulations with mechanical testings to investigate the mechanical behavior of atomic and 3D-printed models of pentadiamond. Pentadiamond is a recently proposed new carbon allotrope, which is composed of a covalent network of pentagonal rings. Our results showed that the stress-strain behavior is almost scale-independent. The stress-strain curves of the 3D-printed structures exhibit three characteristic regions. For low-strain values, this first region presents a non-linear behavior close to zero, followed by a well-defined linear behavior. The second regime is a quasi-plastic one and the third one is densification followed by structural failures (fracture). The Young's modulus values decrease with the number of pores. The deformation mechanism is bending-dominated and different from the layer-by-layer deformation mechanism observed for other 3D-printed structures. They exhibit good energy absorption capabilities, with some structures even outperforming kevlar. Interestingly, considering the Ashby chart, 3D-printed pentadiamond lies almost on the ideal stretch and bending-dominated lines, making them promising materials for energy absorption applications. △ Less

Submitted 20 May, 2021; originally announced May 2021.

Twin Augmented Architectures for Robust Classification of COVID-19 Chest X-Ray Images

Authors: Kartikeya Badola, Sameer Ambekar, Himanshu Pant, Sumit Soman, Anuradha Sural, Rajiv Narang, Suresh Chandra, Jayadeva

Abstract: The gold standard for COVID-19 is RT-PCR, testing facilities for which are limited and not always optimally distributed. Test results are delayed, which impacts treatment. Expert radiologists, one of whom is a co-author, are able to diagnose COVID-19 positivity from Chest X-Rays (CXR) and CT scans, that can facilitate timely treatment. Such diagnosis is particularly valuable in locations lacking r… ▽ More The gold standard for COVID-19 is RT-PCR, testing facilities for which are limited and not always optimally distributed. Test results are delayed, which impacts treatment. Expert radiologists, one of whom is a co-author, are able to diagnose COVID-19 positivity from Chest X-Rays (CXR) and CT scans, that can facilitate timely treatment. Such diagnosis is particularly valuable in locations lacking radiologists with sufficient expertise and familiarity with COVID-19 patients. This paper has two contributions. One, we analyse literature on CXR based COVID-19 diagnosis. We show that popular choices of dataset selection suffer from data homogeneity, leading to misleading results. We compile and analyse a viable benchmark dataset from multiple existing heterogeneous sources. Such a benchmark is important for realistically testing models. Our second contribution relates to learning from imbalanced data. Datasets for COVID X-Ray classification face severe class imbalance, since most subjects are COVID -ve. Twin Support Vector Machines (Twin SVM) and Twin Neural Networks (Twin NN) have, in recent years, emerged as effective ways of handling skewed data. We introduce a state-of-the-art technique, termed as Twin Augmentation, for modifying popular pre-trained deep learning models. Twin Augmentation boosts the performance of a pre-trained deep neural network without requiring re-training. Experiments show, that across a multitude of classifiers, Twin Augmentation is very effective in boosting the performance of given pre-trained model for classification in imbalanced settings. △ Less

Submitted 16 February, 2021; originally announced February 2021.

MSC Class: 68T07

Unsupervised Domain Adaptation for Semantic Segmentation of NIR Images through Generative Latent Search

Authors: Prashant Pandey, Aayush Kumar Tyagi, Sameer Ambekar, Prathosh AP

Abstract: Segmentation of the pixels corresponding to human skin is an essential first step in multiple applications ranging from surveillance to heart-rate estimation from remote-photoplethysmography. However, the existing literature considers the problem only in the visible-range of the EM-spectrum which limits their utility in low or no light settings where the criticality of the application is higher. T… ▽ More Segmentation of the pixels corresponding to human skin is an essential first step in multiple applications ranging from surveillance to heart-rate estimation from remote-photoplethysmography. However, the existing literature considers the problem only in the visible-range of the EM-spectrum which limits their utility in low or no light settings where the criticality of the application is higher. To alleviate this problem, we consider the problem of skin segmentation from the Near-infrared images. However, Deep learning based state-of-the-art segmentation techniques demands large amounts of labelled data that is unavailable for the current problem. Therefore we cast the skin segmentation problem as that of target-independent Unsupervised Domain Adaptation (UDA) where we use the data from the Red-channel of the visible-range to develop skin segmentation algorithm on NIR images. We propose a method for target-independent segmentation where the 'nearest-clone' of a target image in the source domain is searched and used as a proxy in the segmentation network trained only on the source domain. We prove the existence of 'nearest-clone' and propose a method to find it through an optimization algorithm over the latent space of a Deep generative model based on variational inference. We demonstrate the efficacy of the proposed method for NIR skin segmentation over the state-of-the-art UDA segmentation methods on the two newly created skin segmentation datasets in NIR domain despite not having access to the target NIR data. Additionally, we report state-of-the-art results for adaption from Synthia to Cityscapes which is a popular setting in Unsupervised Domain Adaptation for semantic segmentation. The code and datasets are available at https://github.com/ambekarsameer96/GLSS. △ Less

Submitted 17 July, 2020; v1 submitted 15 June, 2020; originally announced June 2020.

Comments: ECCV 2020 [Spotlight]

Zeolite-inspired 3d printed structures with enhanced mechanical properties

Authors: Rushikesh S. Ambekar, Eliezer F. Oliveira, Brijesh Kushwaha, Leonardo D. Machado, Mohammad Sajadi, Ray H. Baughman, Pulickel M. Ajayan, Ajit K. Roy, Douglas S. Galvao, Chandra S. Tiwary

Abstract: Specific strength (strength/density) is a crucial factor while designing high load bearing architecture in areas of aerospace and defence. Strength of the material can be enhanced by blending with high strength component or, by compositing with high strength fillers but both the options has limitations such as at certain load, materials fail due to poor filler and matrix interactions. Therefore, r… ▽ More Specific strength (strength/density) is a crucial factor while designing high load bearing architecture in areas of aerospace and defence. Strength of the material can be enhanced by blending with high strength component or, by compositing with high strength fillers but both the options has limitations such as at certain load, materials fail due to poor filler and matrix interactions. Therefore, researchers are interested in enhancing strength of materials by playing with topology/geometry and therefore nature is best option to mimic for structures whereas, complexity limits nature mimicked structures. In this paper, we have explored Zeolite-inspired structures for load bearing capacity. Zeolite-inspired structure were obtained from molecular dynamics simulation and then fabricated via Fused deposition Modeling. The atomic scale complex topology from simulation is experimentally synthesized using 3D printing. Compressibility of as-fabricated structures was tested in different direction and compared with simulation results. Such complex architecture can be used for ultralight aerospace and automotive parts. △ Less

Submitted 12 February, 2020; v1 submitted 2 February, 2020; originally announced February 2020.

Comments: 24 pages, 6 figures

Journal ref: Additive Manufacturing, 2021, 101628