- Browse by Subject
Browsing by Subject "model"
Now showing 1 - 4 of 4
Results Per Page
Sort Options
Item AI on the Edge with CondenseNeXt: An Efficient Deep Neural Network for Devices with Constrained Computational Resources(2021-08) Kalgaonkar, Priyank B.; El-Sharkawy, Mohamed A.; King, Brian S.; Rizkalla, Maher E.Research work presented within this thesis propose a neoteric variant of deep convolutional neural network architecture, CondenseNeXt, designed specifically for ARM-based embedded computing platforms with constrained computational resources. CondenseNeXt is an improved version of CondenseNet, the baseline architecture whose roots can be traced back to ResNet. CondeseNeXt replaces group convolutions in CondenseNet with depthwise separable convolutions and introduces group-wise pruning, a model compression technique, to prune (remove) redundant and insignificant elements that either are irrelevant or do not affect performance of the network upon disposition. Cardinality, a new dimension to the existing spatial dimensions, and class-balanced focal loss function, a weighting factor inversely proportional to the number of samples, has been incorporated in order to relieve the harsh effects of pruning, into the design of CondenseNeXt’s algorithm. Furthermore, extensive analyses of this novel CNN architecture was performed on three benchmarking image datasets: CIFAR-10, CIFAR-100 and ImageNet by deploying the trained weight on to an ARM-based embedded computing platform: NXP BlueBox 2.0, for real-time image classification. The outputs are observed in real-time in RTMaps Remote Studio’s console to verify the correctness of classes being predicted. CondenseNeXt achieves state-of-the-art image classification performance on three benchmark datasets including CIFAR-10 (4.79% top-1 error), CIFAR-100 (21.98% top-1 error) and ImageNet (7.91% single model, single crop top-5 error), and up to 59.98% reduction in forward FLOPs compared to CondenseNet. CondenseNeXt can also achieve a final trained model size of 2.9 MB, however at the cost of 2.26% in accuracy loss. Thus, performing image classification on ARM-Based computing platforms without requiring a CUDA enabled GPU support, with outstanding efficiency.Item Caenorhabditis elegans as a Model to Study the Molecular and Genetic Mechanisms of Drug Addiction(Elsevier, 2016) Engleman, Eric A.; Katner, Simon N.; Neal-Beliveau, Bethany S.; Department of Psychiatry, IU School of MedicineDrug addiction takes a massive toll on society. Novel animal models are needed to test new treatments and understand the basic mechanisms underlying addiction. Rodent models have identified the neurocircuitry involved in addictive behavior and indicate that rodents possess some of the same neurobiologic mechanisms that mediate addiction in humans. Recent studies indicate that addiction is mechanistically and phylogenetically ancient and many mechanisms that underlie human addiction are also present in invertebrates. The nematode Caenorhabditis elegans has conserved neurobiologic systems with powerful molecular and genetic tools and a rapid rate of development that enables cost-effective translational discovery. Emerging evidence suggests that C. elegans is an excellent model to identify molecular mechanisms that mediate drug-induced behavior and potential targets for medications development for various addictive compounds. C. elegans emit many behaviors that can be easily quantitated including some that involve interactions with the environment. Ethanol (EtOH) is the best-studied drug-of-abuse in C. elegans and at least 50 different genes/targets have been identified as mediating EtOH's effects and polymorphisms in some orthologs in humans are associated with alcohol use disorders. C. elegans has also been shown to display dopamine and cholinergic system-dependent attraction to nicotine and demonstrate preference for cues previously associated with nicotine. Cocaine and methamphetamine have been found to produce dopamine-dependent reward-like behaviors in C. elegans. These behavioral tests in combination with genetic/molecular manipulations have led to the identification of dozens of target genes/systems in C. elegans that mediate drug effects. The one target/gene identified as essential for drug-induced behavioral responses across all drugs of abuse was the cat-2 gene coding for tyrosine hydroxylase, which is consistent with the role of dopamine neurotransmission in human addiction. Overall, C. elegans can be used to model aspects of drug addiction and identify systems and molecular mechanisms that mediate drug effects. The findings are surprisingly consistent with analogous findings in higher-level organisms. Further, model refinement is warranted to improve model validity and increase utility for medications development.Item DESIGN EVALUATION AND DEVELOPMENT OF A VEHICLE PHYSICS MODEL FOR A DRIVER TRAINING SIMULATOR(2017-04-14) Stover, Tyler; Borme, Andrew; Hylton, Peter; Cooney, ElaineAs part of the development of the RLAPS Simulation Software Program (SSP) a vehicle physics model was developed around four subsystems – chassis and suspension, aerodynamics, powertrain, and tires. Tires are the most complex model, and have the most direct impact on the performance and feel of the vehicle model. A complex algorithm governing vehicle physics was presented in a generalized form to guide the programming of the RLAPS SSP. From the generalized algorithm, a practical model was implemented using Unity 3D game creation software (Unity, 2017). The simulation was tested and evaluated against data from numeric lap-time simulation software. Various parameters were opened for tuning to refine the performance and behavior of the vehicle in simulation. The tuned vehicle model performed in such a manner as to exercise the steering, braking, and throttle application skills of drivers using the simulator.Item The Medication Adherence Context and Outcomes Framework Image(2018-10-04) Bartlett Ellis, Rebecca J.; Ruppar, Todd M.Background: Adherence interventions have been largely ineffective, with most taking a "one-size-fits-all” approach without consideration of reasons for nonadherence. While the ABC Taxonomy clarified terminology and identified various outcomes measured along the process continuum, intervention design requires understanding the environments and contexts that contribute to nonadherence. A framework that combines the understanding of environment contextual influences, processes, and outcomes is needed to move forward with approaches to intervention design. Methods: Developed based on theory, practice, and research, the Medication-management and Adherence Contexts and Outcomes (MACO) framework describes the environmental contexts, the processes that occur within the contexts, and how these processes contribute to adherence outcomes. The MACO framework differentiates the processes, defined as medication management, within and across contexts that affect adherence outcomes. Results: Three distinct yet interrelated contexts identified in the MACO framework include 1.) clinic, 2.) pharmacy, and 3.) home. Conclusions: The MACO framework is a useful heuristic to understand at which point people experience problems with managing medications in the medication management continuum. This information can then be used for designing and delivering context-specific interventions and selecting appropriate outcome measures of adherence based on the contexts.