Matching Items (19)

128022-Thumbnail Image.png

Reconciling the differences between a bottom-up and inverse-estimated FFCO2 emissions estimate in a large US urban area

Description

The INFLUX experiment has taken multiple approaches to estimate the carbon dioxide (CO[subscript 2]) flux in a domain centered on the city of Indianapolis, Indiana. One approach, Hestia, uses a

The INFLUX experiment has taken multiple approaches to estimate the carbon dioxide (CO[subscript 2]) flux in a domain centered on the city of Indianapolis, Indiana. One approach, Hestia, uses a bottom-up technique relying on a mixture of activity data, fuel statistics, direct flux measurement and modeling algorithms. A second uses a Bayesian atmospheric inverse approach constrained by atmospheric CO[subscript 2] measurements and the Hestia emissions estimate as a prior CO[subscript 2] flux. The difference in the central estimate of the two approaches comes to 0.94 MtC (an 18.7% difference) over the eight-month period between September 1, 2012 and April 30, 2013, a statistically significant difference at the 2-sigma level. Here we explore possible explanations for this apparent discrepancy in an attempt to reconcile the flux estimates. We focus on two broad categories: 1) biases in the largest of bottom-up flux contributions and 2) missing CO[subscript 2] sources. Though there is some evidence for small biases in the Hestia fossil fuel carbon dioxide (FFCO2) flux estimate as an explanation for the calculated difference, we find more support for missing CO[subscript 2] fluxes, with biological respiration the largest of these. Incorporation of these differences bring the Hestia bottom-up and the INFLUX inversion flux estimates into statistical agreement and are additionally consistent with wintertime measurements of atmospheric [superscript 14]CO[subscript 2]. We conclude that comparison of bottom-up and top-down approaches must consider all flux contributions and highlight the important contribution to urban carbon budgets of animal and biotic respiration. Incorporation of missing CO[subscript 2] fluxes reconciles the bottom-up and inverse-based approach in the INFLUX domain.

Contributors

Agent

Created

Date Created
  • 2017-08-03

128031-Thumbnail Image.png

Optimizing the Spatial Resolution for Urban CO2 Flux Studies Using the Shannon Entropy

Description

The ‘Hestia Project’ uses a bottom-up approach to quantify fossil fuel CO[subscript 2] (FFCO[subscript 2]) emissions spatially at the building/street level and temporally at the hourly level. Hestia FFCO[subscript 2]

The ‘Hestia Project’ uses a bottom-up approach to quantify fossil fuel CO[subscript 2] (FFCO[subscript 2]) emissions spatially at the building/street level and temporally at the hourly level. Hestia FFCO[subscript 2] emissions are provided in the form of a group of sector-specific vector layers with point, line, and polygon sources to support carbon cycle science and climate policy. Application to carbon cycle science, in particular, requires regular gridded data in order to link surface carbon fluxes to atmospheric transport models. However, the heterogeneity and complexity of FFCO[subscript 2] sources within regular grids is sensitive to spatial resolution. From the perspective of a data provider, we need to find a balance between resolution and data volume so that the gridded data product retains the maximum amount of information content while maintaining an efficient data volume. The Shannon entropy determines the minimum bits that are needed to encode an information source and can serve as a metric for the effective information content. In this paper, we present an analysis of the Shannon entropy of gridded FFCO[subscript 2] emissions with varying resolutions in four Hestia study areas, and find: (1) the Shannon entropy increases with smaller grid resolution until it reaches a maximum value (the max-entropy resolution); (2) total emissions (the sum of several sector-specific emission fields) show a finer max-entropy resolution than each of the sector-specific fields; (3) the residential emissions show a finer max-entropy resolution than the commercial emissions; (4) the max-entropy resolution of the onroad emissions grid is closely correlated to the density of the road network. These findings suggest that the Shannon entropy can detect the information effectiveness of the spatial resolution of gridded FFCO[subscript 2] emissions. Hence, the resolution-entropy relationship can be used to assist in determining an appropriate spatial resolution for urban CO[subscript 2] flux studies. We conclude that the optimal spatial resolution for providing Hestia total FFCO[subscript 2] emissions products is centered around 100 m, at which the FFCO[subscript 2] emissions data can not only fully meet the requirement of urban flux integration, but also be effectively used in understanding the relationships between FFCO[subscript 2] emissions and various social-economic variables at the U.S. census block group level.

Contributors

Agent

Created

Date Created
  • 2017-05-19

128660-Thumbnail Image.png

A Sparse Voxel Octree-Based Framework for Computing Solar Radiation Using 3D City Models

Description

An effective three-dimensional (3D) data representation is required to assess the spatial distribution of the photovoltaic potential over urban building roofs and facades using 3D city models. Voxels have long

An effective three-dimensional (3D) data representation is required to assess the spatial distribution of the photovoltaic potential over urban building roofs and facades using 3D city models. Voxels have long been used as a spatial data representation, but practical applications of the voxel representation have been limited compared with rasters in traditional two-dimensional (2D) geographic information systems (GIS). We propose to use sparse voxel octree (SVO) as a data representation to extend the GRASS GIS r.sun solar radiation model from 2D to 3D. The GRASS GIS r.sun model is nested in an SVO-based computing framework. The presented 3D solar radiation computing framework was applied to 3D building groups of different geometric complexities to demonstrate its efficiency and scalability. We presented a method to explicitly compute diffuse shading losses in r.sun, and found that diffuse shading losses can reduce up to 10% of the annual global radiation under clear sky conditions. Hence, diffuse shading losses are of significant importance especially in complex urban environments.

Contributors

Agent

Created

Date Created
  • 2017-03-31

134929-Thumbnail Image.png

Use of cleavable fluorescent antibodies for highly multiplexed single cell in situ protein analysis

Description

The ability to profile proteins allows us to gain a deeper understanding of organization, regulation, and function of different biological systems. Many technologies are currently being used in order to

The ability to profile proteins allows us to gain a deeper understanding of organization, regulation, and function of different biological systems. Many technologies are currently being used in order to accurately perform the protein profiling. Some of these technologies include mass spectrometry, microarray based analysis, and fluorescence microscopy. Deeper analysis of these technologies have demonstrated limitations which have taken away from either the efficiency or the accuracy of the results. The objective of this project was to develop a technology in which highly multiplexed single cell in situ protein analysis can be completed in a comprehensive manner without the loss of the protein targets. This was accomplished in the span of 3 steps which is referred to as the immunofluorescence cycle. Antibodies with attached fluorophores with the help of novel azide-based cleavable linker are used to detect protein targets. Fluorescence imaging and data storage procedures are done on the targets and then the fluorophores are cleaved from the antibodies without the loss of the protein targets. Continuous cycles of the immunofluorescence procedure can help create a comprehensive and quantitative profile of the protein. The development of such a technique will not only help us understand biological systems such as solid tumor, brain tissues, and developing embryos. But it will also play a role in real-world applications such as signaling network analysis, molecular diagnosis and cellular targeted therapies.

Contributors

Agent

Created

Date Created
  • 2016-12

134940-Thumbnail Image.png

Multiplexed single-cell in situ RNA analysis by reiterative hybridization

Description

Currently, quantification of single cell RNA species in their natural contexts is restricted due to the little number of parallel analysis. Through this, we identify a method to increase the

Currently, quantification of single cell RNA species in their natural contexts is restricted due to the little number of parallel analysis. Through this, we identify a method to increase the multiplexing capacity of RNA analysis for single cells in situ. Initially, RNA transcripts are found by using fluorescence in situ hybridization (FISH). Once imaging and data storage is completed, the fluorescence signal is detached through photobleaching. By doing so, the FISH is reinitiated to detect other RNA species residing in the same cell. After reiterative cycles of hybridization, imaging and photobleaching, the identities, positions and copy numbers of a huge amount of varied RNA species can be computed in individual cells in situ. Through this approach, we have evaluated seven different transcripts in single HeLa cells with five reiterative RNA FISH cycles. This method has the ability to detect over 100 varied RNA species in single cells in situ, which can be further applied in studies of systems biology, molecular diagnosis and targeted therapies.

Contributors

Agent

Created

Date Created
  • 2016-12

158066-Thumbnail Image.png

Towards Robust Machine Learning Models for Data Scarcity

Description

Recently, a well-designed and well-trained neural network can yield state-of-the-art results across many domains, including data mining, computer vision, and medical image analysis. But progress has been limited for tasks

Recently, a well-designed and well-trained neural network can yield state-of-the-art results across many domains, including data mining, computer vision, and medical image analysis. But progress has been limited for tasks where labels are difficult or impossible to obtain. This reliance on exhaustive labeling is a critical limitation in the rapid deployment of neural networks. Besides, the current research scales poorly to a large number of unseen concepts and is passively spoon-fed with data and supervision.

To overcome the above data scarcity and generalization issues, in my dissertation, I first propose two unsupervised conventional machine learning algorithms, hyperbolic stochastic coding, and multi-resemble multi-target low-rank coding, to solve the incomplete data and missing label problem. I further introduce a deep multi-domain adaptation network to leverage the power of deep learning by transferring the rich knowledge from a large-amount labeled source dataset. I also invent a novel time-sequence dynamically hierarchical network that adaptively simplifies the network to cope with the scarce data.

To learn a large number of unseen concepts, lifelong machine learning enjoys many advantages, including abstracting knowledge from prior learning and using the experience to help future learning, regardless of how much data is currently available. Incorporating this capability and making it versatile, I propose deep multi-task weight consolidation to accumulate knowledge continuously and significantly reduce data requirements in a variety of domains. Inspired by the recent breakthroughs in automatically learning suitable neural network architectures (AutoML), I develop a nonexpansive AutoML framework to train an online model without the abundance of labeled data. This work automatically expands the network to increase model capability when necessary, then compresses the model to maintain the model efficiency.

In my current ongoing work, I propose an alternative method of supervised learning that does not require direct labels. This could utilize various supervision from an image/object as a target value for supervising the target tasks without labels, and it turns out to be surprisingly effective. The proposed method only requires few-shot labeled data to train, and can self-supervised learn the information it needs and generalize to datasets not seen during training.

Contributors

Agent

Created

Date Created
  • 2020

158746-Thumbnail Image.png

irRotate - Automatic Screen Rotation Based on Face Orientation using Infrared Cameras

Description

This work solves the problem of incorrect rotations while using handheld devices.Two new methods which improve upon previous works are explored. The first method
uses an infrared camera to capture

This work solves the problem of incorrect rotations while using handheld devices.Two new methods which improve upon previous works are explored. The first method
uses an infrared camera to capture and detect the user’s face position and orient the
display accordingly. The second method utilizes gyroscopic and accelerometer data
as input to a machine learning model to classify correct and incorrect rotations.
Experiments show that these new methods achieve an overall success rate of 67%
for the first and 92% for the second which reaches a new high for this performance
category. The paper also discusses logistical and legal reasons for implementing this
feature into an end-user product from a business perspective. Lastly, the monetary
incentive behind a feature like irRotate in a consumer device and explore related
patents is discussed.

Contributors

Agent

Created

Date Created
  • 2020

3D - Patch Based Machine Learning Systems for Alzheimer’s Disease classification via 18F-FDG PET Analysis

Description

Alzheimer’s disease (AD), is a chronic neurodegenerative disease that usually starts slowly and gets worse over time. It is the cause of 60% to 70% of cases of dementia. There

Alzheimer’s disease (AD), is a chronic neurodegenerative disease that usually starts slowly and gets worse over time. It is the cause of 60% to 70% of cases of dementia. There is growing interest in identifying brain image biomarkers that help evaluate AD risk pre-symptomatically. High-dimensional non-linear pattern classification methods have been applied to structural magnetic resonance images (MRI’s) and used to discriminate between clinical groups in Alzheimers progression. Using Fluorodeoxyglucose (FDG) positron emission tomography (PET) as the pre- ferred imaging modality, this thesis develops two independent machine learning based patch analysis methods and uses them to perform six binary classification experiments across different (AD) diagnostic categories. Specifically, features were extracted and learned using dimensionality reduction and dictionary learning & sparse coding by taking overlapping patches in and around the cerebral cortex and using them as fea- tures. Using AdaBoost as the preferred choice of classifier both methods try to utilize 18F-FDG PET as a biological marker in the early diagnosis of Alzheimer’s . Addi- tional we investigate the involvement of rich demographic features (ApoeE3, ApoeE4 and Functional Activities Questionnaires (FAQ)) in classification. The experimental results on Alzheimer’s Disease Neuroimaging initiative (ADNI) dataset demonstrate the effectiveness of both the proposed systems. The use of 18F-FDG PET may offer a new sensitive biomarker and enrich the brain imaging analysis toolset for studying the diagnosis and prognosis of AD.

Contributors

Agent

Created

Date Created
  • 2017

154885-Thumbnail Image.png

A computational approach to relative image aesthetics

Description

Computational visual aesthetics has recently become an active research area. Existing state-of-art methods formulate this as a binary classification task where a given image is predicted to be beautiful or

Computational visual aesthetics has recently become an active research area. Existing state-of-art methods formulate this as a binary classification task where a given image is predicted to be beautiful or not. In many applications such as image retrieval and enhancement, it is more important to rank images based on their aesthetic quality instead of binary-categorizing them. Furthermore, in such applications, it may be possible that all images belong to the same category. Hence determining the aesthetic ranking of the images is more appropriate. To this end, a novel problem of ranking images with respect to their aesthetic quality is formulated in this work. A new data-set of image pairs with relative labels is constructed by carefully selecting images from the popular AVA data-set. Unlike in aesthetics classification, there is no single threshold which would determine the ranking order of the images across the entire data-set.

This problem is attempted using a deep neural network based approach that is trained on image pairs by incorporating principles from relative learning. Results show that such relative training procedure allows the network to rank the images with a higher accuracy than a state-of-art network trained on the same set of images using binary labels. Further analyzing the results show that training a model using the image pairs learnt better aesthetic features than training on same number of individual binary labelled images.

Additionally, an attempt is made at enhancing the performance of the system by incorporating saliency related information. Given an image, humans might fixate their vision on particular parts of the image, which they might be subconsciously intrigued to. I therefore tried to utilize the saliency information both stand-alone as well as in combination with the global and local aesthetic features by performing two separate sets of experiments. In both the cases, a standard saliency model is chosen and the generated saliency maps are convoluted with the images prior to passing them to the network, thus giving higher importance to the salient regions as compared to the remaining. Thus generated saliency-images are either used independently or along with the global and the local features to train the network. Empirical results show that the saliency related aesthetic features might already be learnt by the network as a sub-set of the global features from automatic feature extraction, thus proving the redundancy of the additional saliency module.

Contributors

Agent

Created

Date Created
  • 2016

155457-Thumbnail Image.png

Deep Learning based Classification of FDG-PET Data for Alzheimer's Disease

Description

Alzheimer’s Disease (AD), a neurodegenerative disease is a progressive disease that affects the brain gradually with time and worsens. Reliable and early diagnosis of AD and its prodromal stages (i.e.

Alzheimer’s Disease (AD), a neurodegenerative disease is a progressive disease that affects the brain gradually with time and worsens. Reliable and early diagnosis of AD and its prodromal stages (i.e. Mild Cognitive Impairment(MCI)) is essential. Fluorodeoxyglucose (FDG) positron emission tomography (PET) measures the decline in the regional cerebral metabolic rate for glucose, offering a reliable metabolic biomarker even on presymptomatic AD patients. PET scans provide functional information that is unique and unavailable using other types of imaging. The computational efficacy of FDG-PET data alone, for the classification of various Alzheimer’s Diagnostic categories (AD, MCI (LMCI, EMCI), Control) has not been studied. This serves as motivation to correctly classify the various diagnostic categories using FDG-PET data. Deep learning has recently been applied to the analysis of structural and functional brain imaging data. This thesis is an introduction to a deep learning based classification technique using neural networks with dimensionality reduction techniques to classify the different stages of AD based on FDG-PET image analysis.

This thesis develops a classification method to investigate the performance of FDG-PET as an effective biomarker for Alzheimer's clinical group classification. This involves dimensionality reduction using Probabilistic Principal Component Analysis on max-pooled data and mean-pooled data, followed by a Multilayer Feed Forward Neural Network which performs binary classification. Max pooled features result into better classification performance compared to results on mean pooled features. Additionally, experiments are done to investigate if the addition of important demographic features such as Functional Activities Questionnaire(FAQ), gene information helps improve performance. Classification results indicate that our designed classifiers achieve competitive results, and better with the additional of demographic features.

Contributors

Agent

Created

Date Created
  • 2017