Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “CALL TRAINING”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 19 records

The composite classification problem in optical information processing

Optical pattern recognition allows objects to be recognized from their images and permits their positional parameters to be estimated accurately in real time. The guiding principle behind optical pattern recognition is that a lens focusing a beam of coherent light modulated with an image produces the two-dimensinal Fourier transform of that image. When the resulting output is further transformed by the matched filter corresponding to the original image, one obtains the autocorrelation function of the original image, which has a peak at the origin. Such a device is called an optical correlator and may be used to recognize the locate the image for which it is designed. (From a practical perspective, an approximation to the matched filter must be used since the spatial light modulator (SLM) on which the filter is implemented usually does not allow one to independently control both the magnitude and phase of the filter.) Generally, one is not just concerned with recognizing a single image but is interested in recognizing a variety of rotated and scaled views of a particular image. In order to recognize these different views using an optical correlator, one may select a subset of these views (whose elements are called training images) and then use a composite filter that is designed to produce a correlation peak for each training image. Presumably, these peaks should be sharp and easily distinguishable from the surrounding correlation plane values. In this report we consider two areas of research regarding composite optical correlators. First, we consider the question of how best to choose the training images that are used to design the composite filter. With regard to quantity, the number of training images should be large enough to adequately represent all possible views of the targeted object yet small enough to ensure that the resolution of the filter is not exhausted. As for the images themselves, they should be distinct enough to avoid numerical difficulties yet similar enough to avoid gaps in which certain views of the target will be unrecognized. One method that we introduce to study this problem is called probing and involves the creation of the artificial imagery. The second problem we consider involves the clasification of the composite filter's correlation plane data. In particular, we would like to determine not only whether or not we are viewing a training image, but, in the former case, we would like to determine which training image is being viewed. This second problem is investigated using traditional M-ary hypothesis testing techniques.

Hall, Eric B.↗

Fuel Cycle Research & Development Technical Monthly – May 2012

Management and Integration LANL - Initial coordination meetings with NE and NA-20 were held the first week of May. Additional meetings, to include NA-82 and the new Proliferation and Terrorism Risk Assessment task are scheduled for June. Work package planning as part of the NTD transition is ongoing. Microcalorimetry LANL - Some parts to attempt repair of a malfunctioning water chiller that has been holding up work on the current detector array have arrived. We will attempt to begin running the compressor/cryostat next week to resume work on commissioning the detector array. We have also been working on a backup plan to temporarily move our system to a different lab space. Completed data analysis and prepared a paper summarizing Pu measurements for peer-review journal. Two members of our team presented work at the SORMA West 2012 conference. Completed data analysis and prepared a milestone report summarizing the development of Monte Carlo tools to study systematic errors. Electrochemical Sensor INL - The second sensor test was stopped short due to the work stand-down. Some activities related to this work have resumed, but not all. The sensor used during the second test was removed from the furnace and it showed cracks and later crumbled when being rinsed for sample preparation. The sensor material will be submitted to SEM analysis once work is resumed in the facilities where analysis is performed. A third sensor test has started. Tests with different arrangements will continue and will be oriented based on post-test analysis of the first three sensor tests. Materials with different annealing temperatures will also be prepared for analysis. Lead slowing down spectrometer LANL - Continued work on perturbation method for LSDS analysis. We are running simulations to delineate issues and are checking additional data libraries to see whether there are differences in simulation results. Electrochemical sensor INL - The second sensor test was stopped short due to the work stand-down. Some activities related to this work have resumed, but not all. The sensor used during the second test was removed from the furnace and it showed cracks and later crumbled when being rinsed for sample preparation. The sensor material will be submitted to SEM analysis once work is resumed in the facilities where analysis is performed. A third sensor test has started. Tests with different arrangements will continue and will be oriented based on post-test analysis of the first three sensor tests. Materials with different annealing temperatures will also be prepared for analysis. Fast Neutron Multiplicity analysis INL - The University of Michigan team continues its work on the post-experiment analysis of the validation data from JRC ISPRA, as well as scoping studies using the simulation tool MCNPX-PoliMi. At INL, our new staff member working on this project is now working to model performance capabilities for the fast-neutrons scintillator design concepts. At INL our primary analysis approach is focused on the Feynman mean-to-variance approach. At the University of Michigan, the primary analysis approach is multiplicity coincidence analysis. Fast Neutron Imaging to Quantify Nuclear Materials ORNL - Completed additional measurements of the two 'trainer' configurations as well as calibration measurements with sources on an x-y stage. The imager was shipped to Pantex for an NA-22 obligation at the end of May. Lead Slowing Down Spectrometer LANL - Continued to look into He4 detector. A Swiss company was located that makes He4 detectors and associated electronics. Had a teleconference with them to assess the possible help we can get from them. We are also looking in to constructing one locally. Ongoing perturbation calculations are providing information on the fundamental systematic error limits of LSDS. In order to achieve separating the contribution of Pu and 235U to the signal, there will need to be tight controls on systematic errors. Multi-isotope Process Monitor PNNL - Attended a training call in advanced multivariate classification and analysis methods. These methods were applied to the multivariate analysis of fuel characteristics based on simulated gamma. Work continued on a computational framework to quantify the impact of gamma-ray counting statistical precision on principal component regression (PCR) analysis of a multi-component (i.e. multi-isotope) gammaray spectrum. Tests of the simulation framework are anticipated to begin in early June. Development of a proposal to instrument H-Canyon is being prepared in conjunction with SRNL and the NNSA's NGSI program is ongoing.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Molar Absorptivities of U(VI), U(IV), and Pu(III) in Nitric Acid Solutions of Various Concentrations Relevant to Developing Nuclear Fuel Recycling Flowsheets

Testing of a co-decontamination (CoDCon) tributyl phosphate (TBP) based solvent extraction flowsheet is being performed to optimize conditions for achieving a target U/Pu ratio in the Pu-containing product. The flowsheet is designed to directly produce a mixture of U and Pu with a U/Pu mass ratio of 7/3. The system is monitored in real time using optical spectroscopy, which allows for control of the U/Pu ratio in the aqueous U/Pu nitrate product. On-line spectrophotometric monitoring of the aqueous stream at the reductive stripping step involves determination of U(VI), U(IV) and Pu(III) concentrations and relies on molar absorptivities of these species in nitric acid at a number of wavelengths corresponding to their peaks’ maxima. The magnitudes of these molar absorptivities are a sensitive function of nitric acid concentration and have to be determined as precisely as possible using an off-line spectrophotometry under well controlled conditions. This step (called training set acquisition) is typically performed not in a high contamination area of radiological glovebox where flow-through cells and other on-line equipment are installed but in a less aggressive environment of a radiological fume hood with much better control of stock solutions quality, dilution factors, optical absorbance scale calibration, etc. This is followed by the calibration transfer between the off-line instrument and the instrument deployed during extraction and stripping runs. This paper reports values of molar absorptivities of U(VI), U(IV), and Pu(III) in nitric acid solution of 0.5 M to 4 M concentration and compares them with available technical literature data. Results of off-line spectrophotometric analysis of selected aqueous grab samples from one of CoDCon runs show satisfactory agreement with total uranium and plutonium concentrations determined in the same samples by ICP-MS.

Sinkov, Sergey I.↗

Profile Images and Annotations for Vehicle Re-identification Algorithms (PRIMAVERA)

This dataset contains 636,246 profile images of vehicles representing 13,963 unique vehicles. The data was collected by a set of roadside sensors over the course of three years. Each time a vehicle passed by one of the sensors, a series of images was collected. The images were processed to detect and localize each vehicle, and a license plate reader collocated with the sensor was used to provide a unique ID for the vehicle. Actual license plate numbers have been obfuscated by replacing with an arbitrary numerical ID for each vehicle. After localizing the vehicle in each image, the original RGB image was rotated, scaled, and shifted to produce a new RGB image of size 234x234 pixels such that the outermost two wheels are located at predetermined pixel locations in the image. In this way, all vehicle images are aligned to one another. This registration process occasionally results in a portion of certain vehicles being cutoff at the edges of the image. The dataset has been partitioned into two sets called training and validation. The two partitions no common vehicles, i.e., a vehicle present in one partition is guaranteed not to be present in the other. In this way, an algorithm can be validated against a set of new vehicles that were not seen during the training process. The training set contains 543,926 images from 64,440 vehicle passes representing 11,918 unique vehicles, while the validation set contains 92,320 images from 10,991 vehicle passes representing 2,045 unique vehicles. Vehicle images are organized by directories corresponding to unique vehicles. The file naming scheme is as follows: veh_{vehID}_tr_{passID}_{frameID}_{elevation}_{timeofday}.jpg where {vehID} is the vehicle ID (unique across the entire dataset), {passID} is an identifier for each tracked vehicle pass (unique across the entire dataset), {frameID} is the index of the frame within the given vehicle pass starting at 0, {elevation} is a two-letter string indicating whether the sensor was elevated (el) or at ground-level (gl), and {timeofday} is a two-letter string indicating whether the image was captured during daytime (dt) or nighttime (nt).

image↗

Development and Characterization of Improved NiTiPd High-Temperature Shape-Memory Alloys by Solid-Solution Strengthening and Thermomechanical Processing

The need for compact, solid-state actuation systems for use in the aerospace, automotive, and other transportation industries is currently motivating research in high-temperature shape-memory alloys (HTSMA) with transformation temperatures greater than 100 C. One of the basic high-temperature alloys investigated to fill this need is Ni(19.5)Ti(50.5)Pd30. Initial testing has indicated that this alloy, while having acceptable work characteristics, suffers from significant permanent deformation (or ratcheting) during thermal cycling under load. In an effort to overcome this deficiency, various solid-solution alloying and thermomechanical processing schemes were investigated. Solid-solution strengthening was achieved by substituting 5at% gold or platinum for palladium in Ni(19.5)Ti(50.5)Pd30, the so-called baseline alloy, to strengthen the martensite and austenite phases against slip processes and improve thermomechanical behavior. Tensile properties, work behavior, and dimensional stability during repeated thermal cycling under load for the ternary and quaternary alloys were compared. The relative difference in yield strength between the martensite and austenite phases and the dimensional stability of the alloy were improved by the quaternary additions, while work output was only minimally impacted. The three alloys were also thermomechanically processed by cycling repeatedly through the transformation range under a constant stress. This so-called training process dramatically improved the dimensional stability in these samples and also recovered the slight decrease in work output caused by quaternary alloying. An added benefit of the solid-solution strengthening was maintenance of enhanced dimensional stability of the trained material to higher temperatures compared to the baseline alloy, providing a greater measure of over-temperature capability.

Bigelow, Glen↗

Addressing Human Error in International Space Station Flight Control Teams: Advances in Ground Training for Science Operators

In flight control, as with any human in the loop system, operator error is an inevitable reality. On the International Space Station (ISS) where crew time and physical resources are precious and often irreplaceable, operator errors can result in significant, irreversible consequences. Flight controllers at the Payload Operations Integration Center (POIC) located at NASA's Marshall Space Flight Center (MSFC) in Huntsville, Alabama know this reality well. At the POIC, operator errors can be caused by a variety of factors, from poor hardware or software design to environmental factors such as time pressure or fatigue. The most difficult errors to address, however, are those which result from ineffective teamwork. To address these teamwork errors, trainers at the POIC have drawn best practices from high reliability industries as well as from our sister ISS control center at the Johnson Space Center (JSC) in Houston, Texas, to develop and implement a new training program focused specifically on teamwork skills. This training program, called Team Skills Training, is inspired by modern and historical training programs developed by NASA and is specifically tailored to the needs of the payload operations flight control team at the POIC. The program consists of training for both certified and trainee flight controllers, and covers team skills topics such as situational awareness, leadership, and communication skills. To maximize effectiveness, the program uses novel instructional techniques which extend beyond the classroom to encourage students to apply what they have learned to their day to day work. Minimizing operator errors in flight control is an endeavor which requires constant vigilance and continuous improvement. At the POIC, Team Skills Training is an important step in this journey.

Harris, Samantha S.↗

The relationship of motion sickness susceptibility to learned autonomic control for symptom suppression

Twenty-four men were randomly assigned to four equal groups matched in terms of their Coriolis Sickness Susceptibility Index (CSSI). Two groups of subjects were highly susceptible to motion sickness, and two groups were moderately susceptible. All subjects were given six C551 tests at 5-d intervals. Treatment Groups I (highly susceptible) and II (moderately susceptible) were taught to control their autonomic responses, using a training method called autogenic-feedback training (AFT) before the third, fourth, and fifth CSSI tests. Control groups III (highly susceptible) and IV (moderately susceptible) received no treatment. Results showed that both treatment groups significantly improved performance on CSSI tests after training; neither of the control groups changed significantly. Highly and moderately susceptible subjects in the two treatment groups improved at comparable rates. Highly susceptible control group subjects did not habituate across tests as readily as the moderately susceptible controls.

Cowings, P. S.↗

Space Science Investigation: NASA ISS Stowage Simulator

During this internship the opportunity was granted to work with the Integrated, Graphics, Operations and Analysis Laboratory (IGOAL) team. The main assignment was to create 12 achievement patches for the Space Station training simulator called the "NASA ISS Stowage Training Game." This project was built using previous IGOAL developed software. To accomplish this task, Adobe Photoshop and Adobe Illustrator were used to craft the badges and other elements required. Blender, a 3D modeling software, was used to make the required 3D elements. Blender was a useful tool to make things such as a CTB bag for the "No More Bob" patch which shows a gentleman kicking a CTB bag into the distance. It was also used to pose characters to the positions that was optimal for their patches as in the "Station Sanitation" patch which portrays and astronaut waving on a U.S module on a truck. Adobe Illustrator was the main piece of software for this task. It was used to craft the badges and upload them when they were completed. The style of the badges were flat, meaning that they shouldn't look three dimensional in any way, shape or form. Adobe Photoshop was used when any pictures need brightening and was where the texture for the CTB bag was made. In order for the patches to be ready for the game's next major release, they have to go under some critical reviewing, revising and re-editing to make sure the other artists and the rest of the staff are satisfied with the final products. Many patches were created and revamped to meet the flat setting and incorporate suggestions from the IGOAL team. After the three processes were completed, the badges were implemented into the game (reference fig1 for badges). After a month of designing badges, the finished products were placed into the final game build via the programmers. The art was the final piece in showcasing the latest build to the public for testing. Comments from the testers were often exceptional and the feedback on the badges were outstanding. The latest build will also be tested by astronauts to determine if the product could potentially be incorporated into the astronaut training program. This experience has provided the opportunity to experience the feeling of working with research and development organization. The art skills and programming skills received at the internship will provide and maintain and strong foundation for a career in computer programming and digital media.

Crawford, Gary↗

Damage Detection Using Holography and Interferometry

This paper reviews classical approaches to damage detection using laser holography and interferometry. The paper then details the modern uses of electronic holography and neural-net-processed characteristic patterns to detect structural damage. The design of the neural networks and the preparation of the training sets are discussed. The use of a technique to optimize the training sets, called folding, is explained. Then a training procedure is detailed that uses the holography-measured vibration modes of the undamaged structures to impart damage-detection sensitivity to the neural networks. The inspections of an optical strain gauge mounting plate and an International Space Station cold plate are presented as examples.

Decker, Arthur J.↗

Addressing Human Error in International Space Station Flight Control Teams: Advances in Ground Training for Science Operators

In flight control, as with any human in the loop system, operator error is an inevitable reality. On the International Space Station (ISS) where crew time and physical resources are precious and often irreplaceable, operator errors can result in significant, irreversible consequences. Flight controllers at the Payload Operations Integration Center (POIC) located at NASA’s Marshall Space Flight Center (MSFC) in Huntsville, Alabama know this reality well. At the POIC, operator errors can be caused by a variety of factors, from poor hardware or software design to environmental factors such as time pressure or fatigue. The most difficult errors to address, however, are those which result from ineffective teamwork.Academic research in teamwork has resulted in the identification of many factors which make cross-functional teaming difficult, including leadership, trust building, and communication challenges. These factors, especially when combined with the challenging environmental factors flight control teams must contend with daily, make the goal of minimizing operator errors in payload operations challenging to achieve. To address such teamwork errors, trainers at the POIC have drawn best practices from high reliability industries such as commercial aviation, healthcare, and nuclear power plants, as well as from our sister ISS control center in Houston, Texas, to develop and institute a new training program focused specifically on teamwork skills.This training program, called the Team Skills Curriculum, is based on the concept of Crew Resource Management (CRM) which was developed by NASA in the 1970s for the commercial aviation industry in response to a series of aviation disasters resulting from ineffective teamwork. CRM was later tailored by the Johnson Space Center (JSC) for use in astronaut and flight control training. The result, called Space Flight Resource Management (SFRM) was formally introduced into manned spaceflight training in the late 90s. SFRM has evolved over the years, but the focus has remained on helping operators develop the skills needed to work as part of an effective team. Using these concepts as well as the latest research in cross-functional teaming and data on specific errors occurring at the POIC, trainers in the integrated flight control training branch created a custom training program for both new and certified payload operations specialists.

Harris, Samantha S.↗

Towards a High Fidelity Training Environment for Autonomous Cyber Defense Agents

Cyber defenders are overwhelmed by the frequency and scale of attacks against their networks. This problem will only be exacerbated as attackers leverage AI to automate their workflows. Autonomous cyber defense capabilities could aid defenders by automating operations and adapting dynamically to novel threats. However, existing training environments fall short in areas such as generalization, explainability, scalability, and transferability, making it intractable to train agents that will be effective in real networks. In this paper we take an important step towards creating autonomous cyber defense agents — we present a high fidelity training environment called Cyberwheel that includes both simulation and emulation capabilities. Cyberwheel simplifies customization of the training network and easily allows redefining the agent’s reward function, observation space, and action space to support rapid experimentation of novel approaches to agent design. It also provides visibility into agent behaviors necessary for agent evaluation and sufficient documentation / examples to lower the barrier to entry. As an example use case of Cyberwheel, we present initial results training an autonomous agent to deploy cyber deception strategies in simulation.

Oesch, T↗

GSplit: Scaling Graph Neural Network Training on Large Graphs via Split-Parallelism

Graph neural networks (GNNs), an emerging class of machine learning models for graphs, have gained popularity for their superior performance in various graph analytical tasks. Mini-batch training is commonly used to train GNNs on large graphs, and data parallelism is the standard approach to scale mini-batch training across multiple GPUs. Data parallel approaches contain redundant work as subgraphs sampled by different GPUs contain significant overlap. To address this issue, we introduce a hybrid parallel mini-batch training paradigm called Split parallelism. Split parallelism avoids redundant work by splitting the sampling, loading, and training of each mini-batch across multiple GPUs. Split parallelism, however, introduces communication overheads that can be more than the savings from removing redundant work. We further present a lightweight partitioning algorithm that probabilistically minimizes these overheads. We implement spllit parllelism in GSplit and show that it outperforms state-of-the-art mini-batch training systems like DGL, Quiver, and P3.

Lim, Seung-Hwan [ORNL] (ORCID:0000000194616866)↗

TEACH (Train to Enable/Achieve Culturally Sensitive Healthcare)

Personnel from diverse ethnic and demographic backgrounds come together in both civilian and military healthcare systems, facing diagnoses that at one level are equalizers: coronary disease is coronary disease, breast cancer is breast cancer. Yet the expression of disease in individuals from different backgrounds, individual patient experience of disease as a particular illness, and interactions between patients and providers occurring in any given disease scenario, all vary enormously depending on the fortuity of the equation of "which patient happens to arrive in whose exam room." Previously, providers' absorption of lessons-learned depended on learning as an apprentice would when exposed over time to multiple populations. As a result, and because providers are often thrown into situations where communications falter through inadequate direct patient experience, diversity in medicine remains a training challenge. The questions then become: Can simulation and virtual training environments (VTEs) be deployed to short-track and standardize this sort of random-walk problem? Can we overcome the unevenness of training caused by some providers obtaining the valuable exposure to diverse populations, whereas others are left to "sink or swim"? This paper summarizes developing a computer-based VTE called TEACH (Training to Enable/Achieve Culturally Sensitive Healthcare). TEACH was developed to enhance healthcare providers' skills in delivering culturally sensitive care to African-American women with breast cancer. With an authoring system under development to ensure extensibility, TEACH allows users to role-play in clinical oncology settings with virtual characters who interact on the basis of different combinations of African American sub-cultural beliefs regarding breast cancer. The paper reports on the roll-out and evaluation of the degree to which these interactions allow providers to acquire, practice, and refine culturally appropriate communication skills and to achieve cultural and individual personalization of healthcare in their clinical practices.

Maulitz, Russell↗

Cascade Back-Propagation Learning in Neural Networks

The cascade back-propagation (CBP) algorithm is the basis of a conceptual design for accelerating learning in artificial neural networks. The neural networks would be implemented as analog very-large-scale integrated (VLSI) circuits, and circuits to implement the CBP algorithm would be fabricated on the same VLSI circuit chips with the neural networks. Heretofore, artificial neural networks have learned slowly because it has been necessary to train them via software, for lack of a good on-chip learning technique. The CBP algorithm is an on-chip technique that provides for continuous learning in real time. Artificial neural networks are trained by example: A network is presented with training inputs for which the correct outputs are known, and the algorithm strives to adjust the weights of synaptic connections in the network to make the actual outputs approach the correct outputs. The input data are generally divided into three parts. Two of the parts, called the "training" and "cross-validation" sets, respectively, must be such that the corresponding input/output pairs are known. During training, the cross-validation set enables verification of the status of the input-to-output transformation learned by the network to avoid over-learning. The third part of the data, termed the "test" set, consists of the inputs that are required to be transformed into outputs; this set may or may not include the training set and/or the cross-validation set. Proposed neural-network circuitry for on-chip learning would be divided into two distinct networks; one for training and one for validation. Both networks would share the same synaptic weights.

Duong, Tuan A.↗

The European Safeguards Research and Development Association Addresses Safeguards and Nonproliferation

The renaissance of efforts to expand the use of nuclear energy requires the parallel development of a renewed and more sophisticated work force. Growth in the nuclear sector with high standard of safety, safeguards and security requires skilled staff for design, operations, inspections etc. High-quality nuclear technology educational programs are diminished from past years, and the ability of universities to attract students and to meet future staffing requirements of the nuclear industry is becoming seriously compromised. Thus, education and training in nuclear engineering and sciences is one of the cornerstones for the nuclear sector. Teaching in the nuclear field still seems strongly influenced by national history but it is time to strengthen resources and collaborate. Moreover with the current nuclear security threats it becomes critical that nuclear technology experts master the basic principles not only of safety, but also of nuclear safeguards, nonproliferation and nuclear security. In Europe the European Nuclear Education Network (ENEN) Association has established the certificate 'European Master of Science in Nuclear Engineering (EMSNE)' as the classic nuclear engineering program covering reactor operation and nuclear safety. However, it does not include courses on nonproliferation, safeguards, or dual-use technologies. The lack of education in nuclear safeguards was tackled by the European Safeguards Research and Development Association (ESARDA), through development and implementation of safeguards course modules. Since 2005 the ESARDA Working Group, called the Training and Knowledge Management Working Group, (TKMWG) has worked with the Joint Research Centre (JRC) in Ispra, Italy to organize a Nuclear Safeguards and Nonproliferation course. This five-day course is held each spring at the JRC, and continues to show increasing interest as evidenced by the positive responses of international lecturers and students. The standard set of lectures covers a broad range of subjects, including nuclear material accountancy principles, legal definitions and the regulatory base and inspection tools and techniques. This 60% core part is given by representatives from regulatory bodies (The International Atomic Energy Agency (IAEA), Institute for Radiological Protection and Nuclear Safety, Directorate General for Nuclear Energy and Transport), industry (AREVA, British Nuclear Group), and research (Stockholm University, Hamburg University, Joint Research Centre-Institute of Transuranic Elements, and Joint Research Centre-Institute for the Protection of the Citizen). The remaining part is completed with topical lectures addressed by invited lecturers, such as from Pacific Northwest National Laboratory and the IAEA addressing topics of physical protection, illicit trafficking, the Iraq case study, exercises, including satellite imagery interpretation etc. With this structure of a stable core plus a variable set of invited lectures, the course will remain sustainable and up-to-date. A syllabus provides the students a homogeneous set of information material in nuclear safeguards and nonproliferation matters at the European and international level. In this way, the ESARDA TKMWG aims to contribute to a two-fold scientific-technical and political-juridical education and training.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Biomimetic, Soft-Material Synapse for Neuromorphic Computing: from Device to Network

Neuromorphic computing refers to a variety of brain-inspired computers, devices, and models inspired by the interconnectivity, performance, and energy efficiency of the human brain. Unlike the ubiquitous von Neumann computer architectures with complex processor cores and sequential computation, biological neurons and synapses operate by storing and processing information simultaneously with the capacity of flexible adaptation resulting in massive computational capability with much less power consumption. The search for a synaptic material which can closely imitate bio-synapse has led to an alamethicin-doped, synthetic biomembrane which can emulate key synaptic functions due to generic memristive property enabling learning and computation. This two-terminal, biomolecular memristor, in contrast to its solid-state counterparts, features similar structure, switching mechanism, and ionic transport modality as biological synapses while consuming considerably lower power. In this paper, we outline a methodology for using this biomolecular synapse to build neural networks capable of solving real-world problems. The physical mechanism underlying its volatile memristance is explored followed by the development of a model of this device for circuit simulation. We outline a circuit design technique to integrate this synapse with solid-state neuron circuit for hardware implementation. Based on these results, we develop a high level simulation framework and use a training scheme called Evolutionary Optimization for Neuromorphic System (EONS) to generate networks for solving two problems, namely iris dataset classification and EEG classification task. The small network size and comparable to state-of-the-art accuracy of these preliminary networks show its potential to enhance synaptic functionality in next generation neuromorphic hardware.

Sakib hasan, Md↗

DeepAnnotator

Given a collection of pairs ((A) written material such as reports, phrases, bulleted lists, etc., (B) structured labels called annotations) called the training data, we want to create software to take future written material (A) and correctly predict (B). In this case we have used medical reports and medical coder's structured labels as our training data. The software builds a machine learning model of type deep learning, specifically a convolutional neural network (CNN) using the training data so that when new data arrives in the future the annotations can be predicted.

Christian, JamesB.↗

Out of Distribution Detection with Neural Network Anchoring

This is code to reproduce and build on OOD detection from the paper "Out of Distribution Detection with Neural Network Anchoring". Our goal here is to exploit heteroscedastic temperature scaling as a calibration strategy for out of distribution (OOD) detection. Heteroscedasticity here refers to the fact that the optimal temperature parameter for each sample can be different, as opposed to conventional approaches that use the same value for the entire distribution. To enable this, we propose a new training strategy called anchoring that can estimate appropriate temperature values for each sample, leading to state-of-the-art OOD detection performance across several benchmarks. Using NTK theory, we show that this temperature function estimate is closely linked to the epistemic uncertainty of the classifier, which explains its behavior. In contrast to some of the best-performing OOD detection approaches, our method does not require exposure to additional outlier datasets, custom calibration objectives, or model ensembling. Through empirical studies with different OOD detection settings - far OOD, near OOD, and semantically coherent OOD - we establish a highly effective OOD detection approach.

Thiagarajan, Jayaraman↗