Edward Verenich | Cornell University (original) (raw)

Uploads

Papers by Edward Verenich

Research paper thumbnail of Pulmonary Disease Classification Using Globally Correlated Maximum Likelihood: an Auxiliary Attention mechanism for Convolutional Neural Networks

arXiv (Cornell University), Sep 1, 2021

Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulm... more Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulmonary abnormalities in chest radiographs. Two complementary generalization properties of CNNs, translation invariance and equivariance, are particularly useful in detecting manifested abnormalities associated with pulmonary disease, regardless of their spatial locations within the image. However, these properties also come with the loss of exact spatial information and global relative positions of abnormalities detected in local regions. Global relative positions of such abnormalities may help distinguish similar conditions, such as COVID-19 and viral pneumonia. In such instances, a global attention mechanism is needed, which CNNs do not support in their traditional architectures that aim for generalization afforded by translation invariance and equivariance. Vision Transformers provide a global attention mechanism, but lack translation invariance and equivariance, requiring significantly more training data samples to match generalization of CNNs. To address the loss of spatial information and global relations between features, while preserving the inductive biases of CNNs, we present a novel technique that serves as an auxiliary attention mechanism to existing CNN architectures, in order to extract global correlations between salient features. Impact Statement-We improve sensitivity of Convolutional Neural Networks (CNNs) using an auxiliary global attention mechanism (GCML) that enables CNNs to utilize global spatial information similar to Vision Transformers (ViTs). Our technique retains the benefits of spatial invariance and equivariance inherent to CNNs, while allowing spatial information of features to be used as discriminators. GCML retains these inductive biases in data starved environments, which ViTs lack due their architecture, and hence require significantly more training data to achieve a similar level of generalization. Finally, we show impirically, that GCML improves the sensitivity of standard CNNs when classifying pulmonary conditions in chest X-rays. We provide all associated code, data, and models for reproducibility and improvement through further research.

Research paper thumbnail of The Utility of Feature Reuse: Transfer Learning in Data-Starved Regimes

ArXiv, 2020

The use of transfer learning with deep neural networks has increasingly become widespread for dep... more The use of transfer learning with deep neural networks has increasingly become widespread for deploying well-tested computer vision systems to newer domains, especially those with limited datasets. We describe a transfer learning use case for a domain with a data-starved regime, having fewer than 100 labeled target samples. We evaluate the effectiveness of convolutional feature extraction and fine-tuning of overparameterized models with respect to the size of target training data, as well as their generalization performance on data with covariate shift, or out-of-distribution (OOD) data. Our experiments show that both overparameterization and feature reuse contribute to successful application of transfer learning in training image classifiers in data-starved regimes.

Research paper thumbnail of Hazard Detection in Supermarkets using Deep Learning on the Edge

ArXiv, 2020

Supermarkets need to ensure clean and safe environments for both shoppers and employees. Slips, t... more Supermarkets need to ensure clean and safe environments for both shoppers and employees. Slips, trips, and falls can result in injuries that have a physical as well as financial cost. Timely detection of hazardous conditions such as spilled liquids or fallen items on supermarket floors can reduce the chances of serious injuries. This paper presents EdgeLite, a novel, lightweight deep learning model for easy deployment and inference on resource-constrained devices. We describe the use of EdgeLite on two edge devices for detecting supermarket floor hazards. On a hazard detection dataset that we developed, EdgeLite, when deployed on edge devices, outperformed six state-of-the-art object detection models in terms of accuracy while having comparable memory usage and inference time.

Research paper thumbnail of FlexServe: Deployment of PyTorch Models as Flexible REST Endpoints

ArXiv, 2020

The integration of artificial intelligence capabilities into modern software systems is increasin... more The integration of artificial intelligence capabilities into modern software systems is increasingly being simplified through the use of cloud-based machine learning services and representational state transfer architecture design. However, insufficient information regarding underlying model provenance and the lack of control over model evolution serve as an impediment to the more widespread adoption of these services in many operational environments which have strict security requirements. Furthermore, tools such as TensorFlow Serving allow models to be deployed as RESTful endpoints, but require error-prone transformations for PyTorch models as these dynamic computational graphs. This is in contrast to the static computational graphs of TensorFlow. To enable rapid deployments of PyTorch models without intermediate transformations we have developed FlexServe, a simple library to deploy multi-model ensembles with flexible batching.

Research paper thumbnail of Pulmonary Disease Classification Using Globally Correlated Maximum Likelihood: an Auxiliary Attention mechanism for Convolutional Neural Networks

Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulm... more Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulmonary abnormalities in chest radiographs. Two complementary generalization properties of CNNs, translation invariance and equivariance, are particularly useful in detecting manifested abnormalities associated with pulmonary disease, regardless of their spatial locations within the image. However, these properties also come with the loss of exact spatial information and global relative positions of abnormalities detected in local regions. Global relative positions of such abnormalities may help distinguish similar conditions, such as COVID-19 and viral pneumonia. In such instances, a global attention mechanism is needed, which CNNs do not support in their traditional architectures that aim for generalization afforded by translation invariance and equivariance. Vision Transformers provide a global attention mechanism, but lack translation invariance and equivariance, requiring significantl...

Research paper thumbnail of Mitigating the Class Overlap Problem in Discriminative Localization: COVID-19 and Pneumonia Case Study

Explainable AI Within the Digital Transformation and Cyber Physical Systems, 2021

Research paper thumbnail of Improving Explainability of Image Classification in Scenarios with Class Overlap: Application to COVID-19 and Pneumonia

2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA), 2020

Research paper thumbnail of Vaccination allocation in large dynamic networks

Journal of Big Data, 2017

Event propagation over a network is a complex and frequently studied human phenomena . Historical... more Event propagation over a network is a complex and frequently studied human phenomena . Historical examples of information exchange between humans in a social network, past and present, can be seen during colonial expansion of the British Empire [4], memes passed between friends on any assortment of social networks on the internet [1] and transmission of human or animal pathogens such as the virus H1N1 over airways . In the information age computer networks can mirror these types of information exchange, with event information being passed along from node to node when network neighbors communicate through various network protocols .

Research paper thumbnail of Vaccination allocation in large dynamic networks

Journal of Big Data, 2017

Event propagation over a network is a complex and frequently studied human phenomena . Historical... more Event propagation over a network is a complex and frequently studied human phenomena . Historical examples of information exchange between humans in a social network, past and present, can be seen during colonial expansion of the British Empire [4], memes passed between friends on any assortment of social networks on the internet [1] and transmission of human or animal pathogens such as the virus H1N1 over airways . In the information age computer networks can mirror these types of information exchange, with event information being passed along from node to node when network neighbors communicate through various network protocols .

Research paper thumbnail of Representing and Reasoning with Correlated Effects in Probabilistic Causal Models

uncertainreasoning.com

This paper introduces a new form of knowledge representation for probabilistic Uncertain Causal M... more This paper introduces a new form of knowledge representation for probabilistic Uncertain Causal Models. This representation allows for correlations among effects to be more easily represented than in chain rule based models, such as standard Bayesian Networks (BN's). The importance of effect correlations has been recognized since the early days of computerized causal models, especially in the medical community. Currently, when effect correlations are important, techniques based on the Bahadur expansion are usually used rather than techniques based on Bayesian Networks.

Research paper thumbnail of Bayesian Causal Modeling Extended and Applied to Resource Requirements

Please use the same title listed on the 75TH MORSS Disclosure Form 712 A/B. If the title of the p... more Please use the same title listed on the 75TH MORSS Disclosure Form 712 A/B. If the title of the presentation has changed please list both.) “Bayesian Causal Modeling Extended and Applied to Resource Requirements” ... Email: john.lemmer@rl.af.mil Fax:315 330 7723 ...

Research paper thumbnail of Pulmonary Disease Classification Using Globally Correlated Maximum Likelihood: an Auxiliary Attention mechanism for Convolutional Neural Networks

arXiv (Cornell University), Sep 1, 2021

Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulm... more Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulmonary abnormalities in chest radiographs. Two complementary generalization properties of CNNs, translation invariance and equivariance, are particularly useful in detecting manifested abnormalities associated with pulmonary disease, regardless of their spatial locations within the image. However, these properties also come with the loss of exact spatial information and global relative positions of abnormalities detected in local regions. Global relative positions of such abnormalities may help distinguish similar conditions, such as COVID-19 and viral pneumonia. In such instances, a global attention mechanism is needed, which CNNs do not support in their traditional architectures that aim for generalization afforded by translation invariance and equivariance. Vision Transformers provide a global attention mechanism, but lack translation invariance and equivariance, requiring significantly more training data samples to match generalization of CNNs. To address the loss of spatial information and global relations between features, while preserving the inductive biases of CNNs, we present a novel technique that serves as an auxiliary attention mechanism to existing CNN architectures, in order to extract global correlations between salient features. Impact Statement-We improve sensitivity of Convolutional Neural Networks (CNNs) using an auxiliary global attention mechanism (GCML) that enables CNNs to utilize global spatial information similar to Vision Transformers (ViTs). Our technique retains the benefits of spatial invariance and equivariance inherent to CNNs, while allowing spatial information of features to be used as discriminators. GCML retains these inductive biases in data starved environments, which ViTs lack due their architecture, and hence require significantly more training data to achieve a similar level of generalization. Finally, we show impirically, that GCML improves the sensitivity of standard CNNs when classifying pulmonary conditions in chest X-rays. We provide all associated code, data, and models for reproducibility and improvement through further research.

Research paper thumbnail of The Utility of Feature Reuse: Transfer Learning in Data-Starved Regimes

ArXiv, 2020

The use of transfer learning with deep neural networks has increasingly become widespread for dep... more The use of transfer learning with deep neural networks has increasingly become widespread for deploying well-tested computer vision systems to newer domains, especially those with limited datasets. We describe a transfer learning use case for a domain with a data-starved regime, having fewer than 100 labeled target samples. We evaluate the effectiveness of convolutional feature extraction and fine-tuning of overparameterized models with respect to the size of target training data, as well as their generalization performance on data with covariate shift, or out-of-distribution (OOD) data. Our experiments show that both overparameterization and feature reuse contribute to successful application of transfer learning in training image classifiers in data-starved regimes.

Research paper thumbnail of Hazard Detection in Supermarkets using Deep Learning on the Edge

ArXiv, 2020

Supermarkets need to ensure clean and safe environments for both shoppers and employees. Slips, t... more Supermarkets need to ensure clean and safe environments for both shoppers and employees. Slips, trips, and falls can result in injuries that have a physical as well as financial cost. Timely detection of hazardous conditions such as spilled liquids or fallen items on supermarket floors can reduce the chances of serious injuries. This paper presents EdgeLite, a novel, lightweight deep learning model for easy deployment and inference on resource-constrained devices. We describe the use of EdgeLite on two edge devices for detecting supermarket floor hazards. On a hazard detection dataset that we developed, EdgeLite, when deployed on edge devices, outperformed six state-of-the-art object detection models in terms of accuracy while having comparable memory usage and inference time.

Research paper thumbnail of FlexServe: Deployment of PyTorch Models as Flexible REST Endpoints

ArXiv, 2020

The integration of artificial intelligence capabilities into modern software systems is increasin... more The integration of artificial intelligence capabilities into modern software systems is increasingly being simplified through the use of cloud-based machine learning services and representational state transfer architecture design. However, insufficient information regarding underlying model provenance and the lack of control over model evolution serve as an impediment to the more widespread adoption of these services in many operational environments which have strict security requirements. Furthermore, tools such as TensorFlow Serving allow models to be deployed as RESTful endpoints, but require error-prone transformations for PyTorch models as these dynamic computational graphs. This is in contrast to the static computational graphs of TensorFlow. To enable rapid deployments of PyTorch models without intermediate transformations we have developed FlexServe, a simple library to deploy multi-model ensembles with flexible batching.

Research paper thumbnail of Pulmonary Disease Classification Using Globally Correlated Maximum Likelihood: an Auxiliary Attention mechanism for Convolutional Neural Networks

Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulm... more Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulmonary abnormalities in chest radiographs. Two complementary generalization properties of CNNs, translation invariance and equivariance, are particularly useful in detecting manifested abnormalities associated with pulmonary disease, regardless of their spatial locations within the image. However, these properties also come with the loss of exact spatial information and global relative positions of abnormalities detected in local regions. Global relative positions of such abnormalities may help distinguish similar conditions, such as COVID-19 and viral pneumonia. In such instances, a global attention mechanism is needed, which CNNs do not support in their traditional architectures that aim for generalization afforded by translation invariance and equivariance. Vision Transformers provide a global attention mechanism, but lack translation invariance and equivariance, requiring significantl...

Research paper thumbnail of Mitigating the Class Overlap Problem in Discriminative Localization: COVID-19 and Pneumonia Case Study

Explainable AI Within the Digital Transformation and Cyber Physical Systems, 2021

Research paper thumbnail of Improving Explainability of Image Classification in Scenarios with Class Overlap: Application to COVID-19 and Pneumonia

2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA), 2020

Research paper thumbnail of Vaccination allocation in large dynamic networks

Journal of Big Data, 2017

Event propagation over a network is a complex and frequently studied human phenomena . Historical... more Event propagation over a network is a complex and frequently studied human phenomena . Historical examples of information exchange between humans in a social network, past and present, can be seen during colonial expansion of the British Empire [4], memes passed between friends on any assortment of social networks on the internet [1] and transmission of human or animal pathogens such as the virus H1N1 over airways . In the information age computer networks can mirror these types of information exchange, with event information being passed along from node to node when network neighbors communicate through various network protocols .

Research paper thumbnail of Vaccination allocation in large dynamic networks

Journal of Big Data, 2017

Event propagation over a network is a complex and frequently studied human phenomena . Historical... more Event propagation over a network is a complex and frequently studied human phenomena . Historical examples of information exchange between humans in a social network, past and present, can be seen during colonial expansion of the British Empire [4], memes passed between friends on any assortment of social networks on the internet [1] and transmission of human or animal pathogens such as the virus H1N1 over airways . In the information age computer networks can mirror these types of information exchange, with event information being passed along from node to node when network neighbors communicate through various network protocols .

Research paper thumbnail of Representing and Reasoning with Correlated Effects in Probabilistic Causal Models

uncertainreasoning.com

This paper introduces a new form of knowledge representation for probabilistic Uncertain Causal M... more This paper introduces a new form of knowledge representation for probabilistic Uncertain Causal Models. This representation allows for correlations among effects to be more easily represented than in chain rule based models, such as standard Bayesian Networks (BN's). The importance of effect correlations has been recognized since the early days of computerized causal models, especially in the medical community. Currently, when effect correlations are important, techniques based on the Bahadur expansion are usually used rather than techniques based on Bayesian Networks.

Research paper thumbnail of Bayesian Causal Modeling Extended and Applied to Resource Requirements

Please use the same title listed on the 75TH MORSS Disclosure Form 712 A/B. If the title of the p... more Please use the same title listed on the 75TH MORSS Disclosure Form 712 A/B. If the title of the presentation has changed please list both.) “Bayesian Causal Modeling Extended and Applied to Resource Requirements” ... Email: john.lemmer@rl.af.mil Fax:315 330 7723 ...