Dubey's Research: Source Separation from Neurally Inspired Processing

Mohit Dubey, a New Mexico Consortium (NMC) Research Scientist working in the field of neuromorphic computing, has been making strides in his research on Source Separation from Neurally Inspired Processing (SSNIP). This research is a collaboration with Garrett Kenyon and Austin Thresher, also of the NMC.

SSNIP is a neutrally-inspired method of separating bass, drums, vocals and other instruments from sparse encodings of phase-rich Constant-Q representations of stereo musical data.
In this research, sparse encodings are generated from learned features that are tonotopic and divided into spectrally and temporally convolutional dictionaries.
This method is inspired by the hemispheric lateralization of human auditory cortex, the section of the brain that processes information received through hearing.
As opposed to previous approaches, Dubey utilizes different methods to achieve his results. First, he uses a logarithmic frequency representation. He also preserves the phase information of the original sound. This is because the phase problem is the problem of loss of information concerning the phase that can occur when making a physical measurement. Third, he uses multiple encoding dictionaries to capture both fine spectral and temporal information. Last, he removes the noise from the original sources in order to obtain the best results possible.
At this point in time, this exciting neutrally-inspired research is only being applied to musical data. Dubey and his team aim to expand their approach in the future to separate sources in speech, biomedical (EEG, fMRI, etc.) and other types of data.



Publication: Oedometric Small-Angle Neutron Scattering

Rex Hjelm of the New Mexico Consortium has a new publication accepted by Environmental Science and Technology titled “Oedometric Small-Angle Neutron Scattering: In Situ Nano-Pore Structure During Bentonite Consolidation and Swelling in CO2 Environment”.

In this paper, results of oedometric consolidation experiments linked with small-angle neutron scattering (SANS) measurements are presented, using SWy-2 Wyoming bentonite clay in dry and water-bearing N2 and CO2 atmospheres.

Oedometric SANS involves deforming a porous sample under uniaxial strain conditions using applied axial force and internal pore pressure control, and combines with SANS for in situ observation of pore structure evolution and interaction.

Scattering from both the interlayer (clay intra-aggregate) and the free (inter-aggregate) pores is observed. This shows decreasing pore size with dry consolidation and interactions between interlayer and free pore types with swelling and consolidation.

Introducing dry liquid CO2 at zero effective stress (axial stress minus pore pressure) produces large shifts in interlayer scatterers, but is reversible back to pre-CO2 levels upon decreasing pore pressure and increasing effective stress. On the other hand, introduction of wet liquid CO2 produces large, but irreversible changes in interlayer scatterers, which are interpreted to be the combined result of CO2 and H2O intercalation under hydrostatic conditions, but which diminish with application of effective pressure and consolidation to higher bentonite dry densities.

Consideration of CO2 intercalation in smectite-bearing CO2 caprocks needs to include effects of both water and nonhydrostatic stress.

To learn more see:

Dewers, T., Heath, J., Bryan, C., Mang, J., Hjelm, R., Ding, M. and Taylor, M., "Oedometric Small-Angle Neutron Scattering: In Situ Nano-Pore Structure During Bentonite Consolidation and Swelling in CO2 Environments", Environmental Science & Technology, (2018, in press).

Manore's Research: How Environment Affects Infectious Diseases


LANL Staff Scientist and NMC Affiliate Dr. Carrie A. Manore is researching how changes in the environment affect the risks for emerging infectious diseases, in particular zoonotic and vector-borne diseases. 

Manore, and her team, are creating a hierarchy of multi-scale, integrative models which incorporate nonlinear systems of differential equations, ecological networks, and agent-based models to provide a framework for understanding and mitigating the risk of emerging infectious diseases in a spatially heterogeneous environment.

Manore’s research focuses on a few emerging or potentially emerging infectious diseases in the United States, including West Nile virus, dengue, Rift Valley fever, hantavirus, Zika and chikungunya, with an emphasis on how human mobility, climate, and disturbance of the environment modify risk. Her team extended their work to Zika modeling once the emerging outbreak in the Americas started, publishing a paper on Zika risk in the Eastern United States and estimating the true outbreak sizes in several countries in South and Central America.

Manore is now working with collaborators on extending the true outbreak size estimates to more countries and estimating the number of children with birth defects as a result of the Zika outbreak. Manore’s team is also collaborating with LANL and Tulane to use remote sensing, internet, and socio-economic data to predict and forecast risk in Brazil of dengue, chikungunya, and Zika at the municipality level.

This research aims to inform how emerging disease risk can be mitigated by sustainable management of the environment, including better planning of urban and agricultural expansion, more effective and efficient public health strategies such as use of new technology, increased surveillance and data sharing, and informed choices about human manipulation of the environment and the importance of biodiversity. The ultimate goal is to better inform policy makers about these new and emerging risks.

This work is funded by an award from the NSF Science, Engineering and Education for Sustainability Fellows (SEES Fellows Program). 

Van de Sompel's Research: Archiving Online Scholarly Material

Herbert Van de Sompel, a scientist and Los Alamos National Laboratory (LANL) and a research affiliate at the New Mexico Consortium (NMC), is concerned about the problem of how to archive scholarly materials. This has become an issue due to the flood of scholarly articles and artifacts published daily online as well as the fact that the nature of scholarly communication is changing. Researchers are communicating in different ways and on different platforms than in the past. Researchers now often use online portals, often general purpose and not dedicated to scholarly work, that specify different topics such as assessment, discovery, analysis, writing, publication and outreach.

While traditional scholarly resources like journal articles do have archival efforts in place, the broad variety of other types of web-based scholarly objects are largely neglected when archiving is concerned. These scholarly materials can include software, workflows, presentations, video recording of experiments, blog posts that dig deep in to specific aspects of research, project websites, etc. Van de Sompel and colleagues call these neglected online materials the “scholarly orphans”.

These scholarly orphans are regularly referenced in scientific papers. But since the web platforms that host them may disappear and since they are not archived, over time, looking up the referenced materials becomes impossible. So the question is how to archive them.
Van de Sompel is collaborating with Michael L. Nelson of Old Dominion University, Martin Klein of LANL and Harihar Shankar of both LANL and the NMC on a new Andrew W. Mellon funded project to explore how these scholarly orphans can be archived.
In this research, they are exploring the feasibility of an approach that automatically monitors a researcher's account in various web portals and returns information about newly added or updated artifacts. The web location (URL) of such artifacts is then passed on to a web crawler that captures artifacts and associated descriptive metadata, and deposits those in a web archive. The approach assumes knowledge of a researcher's identity in various web portals and an institutional process that operates the monitoring, capturing, and archiving processes. Building on the Memento and Robust Links work that Van de Sompel and colleagues did previously, access to artifacts referenced in scientific papers can automatically be routed to copies in web archives.

To learn more read Van de Sompel’s article: Discovering Scholarly Orphans using ORCID

Radial transport of radiation belt electrons in kinetic field-line resonances

August 2017, Geophysical Research Letters

C. C. Chaston, J. W. Bonnell, J. R. Wygant, G. D. Reeves, D. N. Baker, D. B. Melrose, Iver H. Cairns

A representative case study from the Van Allen Probes during a geomagnetic storm recovery phase reveals enhanced electron fluxes at intermediate pitch angles over energies from ~100 keV to 5 MeV coincident with broadband low-frequency electromagnetic waves.

The statistical properties of these waves are used to build a model for radial diffusion via drift-bounce resonances in kinetic Alfvén eigenmodes/kinetic field-line resonances. Estimated diffusion coefficients indicate timescales for radial transport on the order of hours in storm time events at energies from <100 keV to MeVs over equatorial pitch angles from the edge of the loss cone to nearly perpendicular to the geomagnetic field.

The correlation of kinetic resonances with electron depletions and enhancements during storm main phase and recovery, and the rapid diffusion these waves drive, suggests that they may modulate the outer radiation belt.

To summarize, these researchers have found a non-linear interaction that is extremely efficient at transporting electrons radially (inward or outward). Researchers have looked at these “drift-bounce resonances” before, but by looking at the electric field in the electron’s frame of reference, it has been found that the interaction is much stronger than anyone previously thought.


© 2018 New Mexico Consortium