183 research outputs found
Design Space Exploration of a Spiking LSM Classifier for RADAR applications
No full textRecent trends in platforms for the consumer market increased the need for low-power and reliable classification engines. Spiking Neural Network (SNN) is a new technology that promises to deliver 4 orders of magnitude more performance per watt than competing solutions. Moreover, the adoption of RADAR for gesture detection provides higher reliability compared to image sensors. However, no accepted topology for a temporal SNN classifier focusing on RADAR data exists. In addition, previous research did not account for several design limitations necessary to export the design in analog hardware. In this work, we explore the possible SNN topologies and propose a Liquid State Machine (LSM) with fully-supervised readout, suitable to be exported to a mixed-signal neuron array. A complete parametric model of the architecture and learning rule has been implemented in a simulation environment. Following, the design space was explored in search for the optimal operating region. By analysing the results, we: (i) highlight and explain the effects of several parameters and the trade-off between accuracy and power consumption; (ii) emphasize the need for a good balance between global excitation and inhibition in the LSM; (iii) suggest that the limitations of the proposed design point to the importance of an adequate feature extraction for a stable LSM behaviour and to the unpredictable nature of the SNN backpropagation algorithm, caused by the non differentiability of the spike signals.Electrical Engineering | Embedded System
A new scenario of turbulence theory and an application to pipe turbulence
No full textA new general scenario of turbulence theory is proposed and applied to pipe-flow turbulence. The theory supports tranverse traveling waves which suffer energy loss by internal friction (Joule-like effect). Its predictions are consistent with some characteristic features of pipe turbulence found in recent experiments: (i) existence of traveling waves and their important function and (ii) existence of two large scales (LSM and VLSM), both in turbulent state. In fact, predicted waves are characterized by two scales (wave-length and damping-length). Bulk energy dissipation is expressed unexpectedly in a form analogous to a model of eddy-viscosity
Elastic least-squares migration for quantitative reflection imaging of fracture compliances
No full textTo quantitatively image fractures with high resolution, we have developed an elastic least-squares migration (LSM) algorithm coupled with linear-slip theory, which accurately addresses seismic wave interaction with thin structures. We derive a linearized waveform inversion using the Born approximation to the boundary integral equation for scattered waves, including linear-slip interfaces for P-SV and SH wavefields. Numerical modeling tests assuming a laboratory-scale fracture where a 20 cm long fracture is illuminated by waves with a 50 kHz center frequency show that our LSM successfully estimates fracture compliances. Furthermore, due to the presence of coupling compliances at the fracture, the results using our LSM show better images than those using the conventional LSM estimating the Lamé constants. We also numerically illustrate that our LSM can be successfully applied to dipole acoustic borehole logging data with 3 kHz center frequency for single-well reflection imaging of a 10 m long, dipping fracture embedded in a random background. Finally, we apply LSM to laboratory experimental data, measuring PP reflections from a fluid-filled fracture. We confirm that the estimated fracture compliances correspond well to those estimated by earlier amplitude variation with offset inversion. Furthermore, the LSM resolves the spatially varying fracture compliances due to local filling of water in the fracture. Because the linear-slip theory can be applied to thin structures in a wide range of scales, high-resolution imaging results and estimated fracture compliance distributions will be crucial to further address small-scale properties at fractures, joints, and geologic faults.Accepted author manuscriptApplied Geophysics and PetrophysicsImPhys/Medical Imagin
Quantitative fracture imaging using least-squares migration and linear-slip model: theory and application to single-well reflection imaging
No full textCharacterizing subsurface fractures is a key to developing hydrocarbon and geothermal fields, as well as providing fundamental information on fracture system relevant to regional seismotectonics. Seismic characterization of fractures has generally been based on the effective medium theory, which considers seismically invisible small fractures. Therefore, there is a considerable scale gap between the fracture properties obtained by seismic methods and those from borehole logging. Recent studies of single-well reflection imaging using acoustic borehole logging data show the potential of filling the scale gap by providing fracture properties around the borehole up to a few tens of meter away from the borehole location. In the context of reflection imaging of individual fractures, in this study, we develop least-squares migration (LSM) coupled with linear-slip model. LSM solves the linearized waveform inversion to provide high-resolution quantitative images. Linear-slip model can describe wave reflection at a fracture accurately. We show numerical modelling examples of the proposed approach considering a vertical fracture with coupling compliances, and acoustic dipole measurements of a dipping fracture embedded in a background random velocity distribution. The results show that the proposed LSM provides higher resolution images than reverse-time migration, and more accurate images than the conventional LSM without linear-slip model.Accepted author manuscriptApplied Geophysics and PetrophysicsImPhys/Acoustical Wavefield Imagin
Surface discontinuity modelling by LSM through patch adaptation and use of edges
No full textThis paper handles the aspect of least squares matching (LSM) for DSM generation and focuses mainly on the exploitation of edge information in the least squares approach. The mathematical model of LSM employs an affine transformation to model geometrical distortions between the template and the search images. However, the model, when used for single points that lie close or on edges does not suffice to model discontinuities, especially when patch size is large. To alleviate the above-mentioned problem, LSM is firstly extended to edgels using an approximation of the rotation between search image and template, computed from a signal analysis. The signal analysis is performed by computing the signal ellipses from the template and patch grey levels. Furthermore, the template patch can be pre-rotated and its dimensions can be adjusted, so that the patch becomes a thin ribbon along the edge. Secondly, the model is extended to edge features with topological attributes (edges and vertices with link information). The patch size is adapted to include a whole edge and a non-uniform weighting scheme is utilised in the normal equations of LSM, giving a very low weight to all pixels outside the given edge. After matching the edge middle points, the edge is divided in two segments and the procedure is repeated until the edge segments fall below a certain length threshold. The final stage is the matching of the edge end points, where other edges having the same end point are also taken into account. The proposed techniques are compared to each other and to standard LSM. An evaluation of the methods is done with ADS40 imagery. As it is shown, they lead to less discontinuity smoothing and increase of the success rate and accuracy. * Corresponding author. 1
Lethe: a tunable delete-aware LSM engine (updated version)
No full textData-intensive applications fueled the evolution of log structured merge (LSM) based key-value engines that employ the out-of-place paradigm to support high ingestion rates with low read/write interference. These benefits, however, come at the cost of treating deletes as a second-class citizen. A delete inserts a tombstone that invalidates older instances of the deleted key. State-of-the-art LSM engines do not provide guarantees as to how fast a tombstone will propagate to persist the deletion. Further, LSM engines only support deletion on the sort key. To delete on another attribute (e.g., timestamp), the entire tree is read and re-written. We highlight that fast persistent deletion without affecting read performance is key to support: (i) streaming systems operating on a window of data, (ii) privacy with latency guarantees on the right-to-be-forgotten, and (iii) en masse cloud deployment of data systems that makes storage a precious resource. To address these challenges, in this paper, we build a new key-value storage engine, Lethe, that uses a very small amount of additional metadata, a set of new delete-aware compaction policies, and a new physical data layout that weaves the sort and the delete key order. We show that Lethe supports any user-defined threshold for the delete persistence latency offering higher read throughput (1.17-1.4×) and lower space amplification (2.1-9.8×), with a modest increase in write amplification (between 4% and 25%). In addition, Lethe supports efficient range deletes on a secondary delete key by dropping entire data pages without sacrificing read performance nor employing a costly full tree merge.First author draf
OpenKV: LSM-tree-based key-value store for open-channel SSD
No full textLog-structured merge (LSM) tree-based key-value stores, such as LevelDB and RocksDB, have seen great adoption in industry due to their high write speed. However, one major issue with LSM-based databases is the high write amplification. The root cause of this problem is the LSM tree structure that demands each level to be completely sorted. In this work, we propose OpenKV, a novel key-value store for open-channel SSD that achieves very low write amplification with good read performance. We propose a design with partially sorted levels with lazy compaction to reduce write amplification, and we have designed a central lookup table based on the cuckoo filter to utilize the open-channel SSD's direct page-level access. In our evaluation we show that compared to LevelDB, our design can reduce write traffic by 2.6x to 3.5x while improving random read performance by up to 1.8x.Submission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2021-12-01The student, Xiaohao Wang, accepted the attached license on 2019-11-19 at 14:35.The student, Xiaohao Wang, submitted this Thesis for approval on 2019-11-19 at 14:51.This Thesis was approved for publication on 2019-11-19 at 16:03.DSpace SAF Submission Ingestion Package generated from Vireo submission #14566 on 2020-02-28 at 17:22:25Made available in DSpace on 2020-03-02T22:12:19Z (GMT). No. of bitstreams: 3
WANG-THESIS-2019.pdf: 988208 bytes, checksum: e0f934a9b5fd42ca5c4b5e498e44f68c (MD5)
Master Thesis.zip: 1031677 bytes, checksum: 069c09f13858e68ba0a6f7c3a69f3123 (MD5)
LICENSE.txt: 4209 bytes, checksum: 9ff8f45ee449fb9b76319fa72af5c558 (MD5)
Previous issue date: 2019-11-19Embargo set by: Seth Robbins for item 113890
Lift date: 2022-03-02T22:12:26Z
Reason: Author requested U of Illinois access only (OA after 2yrs) in Vireo ETD systemEmbargo set by: Seth Robbins for item 113890
Lift date: 2022-03-02T22:15:21Z
Reason: Author requested U of Illinois access only (OA after 2yrs) in Vireo ETD systemEmbargo set by: Seth Robbins for item 113890
Lift date: 2022-03-02T22:18:25Z
Reason: Author requested U of Illinois access only (OA after 2yrs) in Vireo ETD systemU of I Only Restriction Lifted for Item 113890 on 2022-03-03T10:15:08Z
A seismic vertical vibrator driven by linear synchronous motors
No full textA linear synchronous motor (LSM) is an electric motor that can produce large controllable forces and is therefore suitable as a driving engine for a seismic vibrator. This motor consists of two independent elements, a magnet track and a coil track, allowing practically unlimited motor displacements. This makes the LSM very suitable for expanding the source frequency band to the lower frequencies in which larger strokes are needed. In contrast to hydraulic engines, the LSM performs equally well over the whole frequency range, making possible a smaller amount of signal distortion, especially at the low frequencies. To find the feasibility of an LSM-driven vibrator, we successfully designed and built a multi-LSM prototype vibrator of some 1200 kg. We addressed the synchronization between the individual motor tracks and the different motors. To lower the energy consumption, a spring mechanism was implemented that delivered the force needed to lift the vibrator mass to its neutral position. The resonance belonging to this spring mechanism was successfully suppressed with the help of a position feedback control that also suppressed the temperature effects. The seismic data acquired in the field tests proved that the prototype LSM vibrator acted very well as a seismic source. It has no trouble generating pseudorandom sweeps, and even given its limited size, it generated signals within the low-frequency regime, down to 2 Hz, rather easily.Geoscience & EngineeringCivil Engineering and Geoscience
The effect of temperature fluctuations on the spread of a buoyant plume
No full textEmissions from many natural and anthropogenic sources are hot compared with the surrounding ambient air. Such buoyancy effects cause the emitted plume to rise, increasing the effective source height and significantly decreasing the maximum ground level concentrations (in the vicinity of the source). A major aspect that distinguishes buoyant and passive dispersion is that buoyant fluid particles create their own turbulence and hence exchange processes between the plume and its environment need to be accounted for. The inclusion of plume rise in Lagrangian stochastic models (LSMs) of turbulent dispersion has been considered by many authors but the interaction of the buoyant plume with the environment (by means of entrainment) is difficult to model in a Lagrangian framework. Webster and Thomson [8] formulated a hybrid model in which the mean flow is calculated from a simple plume model and the fluctuations of velocity are calculated using an LSM. They model the effect of turbulence generated by the plume by an additional random increment to the position of a particle. Here, instead of including this extra term, we add a stochastic differential equation (SDE) for the temperature fluctuations suitably coupled with the SDE for the velocity fluctuations. The interaction of temperature and velocity fluctuations, directly related to the turbulence within the plume, determines the plume’s spread. The results of the model are compared with large-eddy simulation (LES) of buoyant plumes in a uniform crosswind and also with the plume generated by the explosion and fire at the Buncefield oil depot in 2005 using realistic profiles of the wind speed and direction and thermal stratification
Reduction of computing time for least-squares migration based on the Helmholtz equation by graphics processing units
No full textIn geophysical applications, the interest in least-squares migration (LSM) as an imaging algorithm is increasing due to the demand for more accurate solutions and the development of high-performance computing. The computational engine of LSM in this work is the numerical solution of the 3D Helmholtz equation in the frequency domain. The Helmholtz solver is Bi-CGSTAB preconditioned with the shifted Laplace matrix-dependent multigrid method. In this paper, an efficient LSM algorithm is presented using several enhancements. First of all, a frequency decimation approach is introduced that makes use of redundant information present in the data. It leads to a speedup of LSM, whereas the impact on accuracy is kept minimal. Secondly, a new matrix storage format Very Compressed Row Storage (VCRS) is presented. It not only reduces the size of the stored matrix by a certain factor but also increases the efficiency of the matrix-vector computations. The effects of lossless and lossy compression with a proper choice of the compression parameters are positive. Thirdly, we accelerate the LSM engine by graphics cards (GPUs). A GPU is used as an accelerator, where the data is partially transferred to a GPU to execute a set of operations or as a replacement, where the complete data is stored in the GPU memory. We demonstrate that using the GPU as a replacement leads to higher speedups and allows us to solve larger problem sizes. Summarizing the effects of each improvement, the resulting speedup can be at least an order of magnitude compared to the original LSM method.Delft Institute of Applied MathematicsElectrical Engineering, Mathematics and Computer Scienc
- …
