Direct classification from compressively sensed images via deep Boltzmann machine

Published on Nov 1, 2016 in ASILOMAR (Asilomar Conference on Signals, Systems and Computers)
· DOI :10.1109/ACSSC.2016.7869080
Henry Braun6
Estimated H-index: 6
Pavan Turaga24
Estimated H-index: 24
+ 1 AuthorsCihan Tepedelenlioglu26
Estimated H-index: 26
We examine a potential technique of performing a classification task based on compressively sensed (CS) data, skipping a computationally expensive reconstruction step. A deep Boltzmann machine is trained on a compressive representation of MNIST handwritten digit data, using a random orthoprojector sensing matrix. The network is first pre-trained on uncompressed data in order to learn the structure of the dataset. The outer network layers are then optimized using backpropagation. We find this approach achieves a 1.21% test data error rate at a sensing rate of 0.4, compared to a 0.99% error rate for non-compressive data.
  • References (16)
  • Citations (3)
📖 Papers frequently viewed together
2013ICICS: International Conference on Information and Communication Security
7 Authors (Aihua Yu, ..., Beiping Hou)
1 Citations
13 Citations
3 Citations
78% of Scinapse members use related papers. After signing in, all features are FREE.
#1Yun Li (Rice University)H-Index: 5
#2Chinmay Hegde (MIT: Massachusetts Institute of Technology)H-Index: 17
Last. Kevin F. Kelly (Rice University)H-Index: 27
view all 5 authors...
Given its importance in a wide variety of machine vision applications, extending high-speed object detection and recognition beyond the visible spectrum in a cost-effective manner presents a significant technological challenge. As a step in this direction, we developed a novel approach for target image classification using a compressive sensing architecture. Here we report the first implementation of this approach utilizing the compressive single-pixel camera system. The core of our approach res...
9 CitationsSource
May 1, 2014 in ICASSP (International Conference on Acoustics, Speech, and Signal Processing)
#1Henry Braun (ASU: Arizona State University)H-Index: 6
#2Pavan Turaga (ASU: Arizona State University)H-Index: 24
Last. Andreas Spanias (ASU: Arizona State University)H-Index: 28
view all 3 authors...
The compressive sensing paradigm holds promise for more cost-effective imaging outside of the visible range, particularly in infrared wavelengths. However, the process of reconstructing compressively sensed images remains computationally expensive. The proof-of-concept tracker described here uses a particle filter with a likelihood update based on a “smashed filter” which estimates correlation directly, avoiding the reconstruction step. This approach leads to increased noise in correlation estim...
2 CitationsSource
#1Jeremy Vila (OSU: Ohio State University)H-Index: 8
#2Philip Schniter (OSU: Ohio State University)H-Index: 40
When recovering a sparse signal from noisy compressive linear measurements, the distribution of the signal's non-zero coefficients can have a profound effect on recovery mean-squared error (MSE). If this distribution was a priori known, then one could use computationally efficient approximate message passing (AMP) techniques for nearly minimum MSE (MMSE) recovery. In practice, however, the distribution is unknown, motivating the use of robust algorithms like LASSO-which is nearly minimax optimal...
224 CitationsSource
May 1, 2013 in ICASSP (International Conference on Acoustics, Speech, and Signal Processing)
#1Henry Braun (ASU: Arizona State University)H-Index: 6
#2Pavan Turaga (ASU: Arizona State University)H-Index: 24
Last. Andreas Spanias (ASU: Arizona State University)H-Index: 28
view all 4 authors...
Although considerable effort has been devoted to the problem of reconstructing compressively sensed video, no existing algorithm achieves results comparable to commonly available video compression methods such as H.264. One possible avenue for improving compressively sensed video reconstruction is the use of optical flow information. Current efforts reported in the literature have not fully utilized optical flow information, instead focusing on limited cases such as stationary backgrounds with s...
6 CitationsSource
Apr 1, 2012 in ICCP (International Conference on Computational Photography)
#1Aswin C. Sankaranarayanan (Rice University)H-Index: 21
#2Christoph Studer (Rice University)H-Index: 34
Last. Richard G. Baraniuk (Rice University)H-Index: 87
view all 3 authors...
Compressive sensing (CS)-based spatial-multiplexing cameras (SMCs) sample a scene through a series of coded projections using a spatial light modulator and a few optical sensor elements. SMC architectures are particularly useful when imaging at wavelengths for which full-frame sensors are too cumbersome or expensive. While existing recovery algorithms for SMCs perform well for static images, they typically fail for time-varying scenes (videos). In this paper, we propose a novel CS multi-scale vi...
114 CitationsSource
Jan 1, 2011 in NeurIPS (Neural Information Processing Systems)
#1Andrew E. Waters (Rice University)H-Index: 7
#2Aswin C. Sankaranarayanan (Rice University)H-Index: 21
Last. Richard G. Baraniuk (Rice University)H-Index: 87
view all 3 authors...
We consider the problem of recovering a matrix M that is the sum of a low-rank matrix L and a sparse matrix S from a small set of linear measurements of the form y = A(M)= A(L + S). This model subsumes three important classes of signal recovery problems: compressive sensing, affine rank minimization, and robust principal component analysis. We propose a natural optimization problem for signal recovery under this model and develop a new greedy algorithm called SpaRCS to solve it. Empirically, Spa...
162 Citations
#1Bob L. Sturm (AAU Cph: Aalborg University – Copenhagen)H-Index: 15
#2Mads Græsbøll Christensen (AAU: Aalborg University)H-Index: 23
Last. Rémi Gribonval (IRIA: French Institute for Research in Computer Science and Automation)H-Index: 44
view all 3 authors...
The pure greedy algorithms matching pursuit (MP) and complementary MP (CompMP) are extremely computationally simple, but can perform poorly in solving the linear inverse problems posed by the recovery of compressively sampled sparse signals. We show that by applying a cyclic minimization principle, the performance of both are significantly improved while remaining computationally simple. Our simulations show that while MP and CompMP may not be competitive with state-of-the-art recovery algorithm...
15 CitationsSource
Jul 1, 2011 in ISIT (International Symposium on Information Theory)
#1Sundeep Rangan (NYU: New York University)H-Index: 44
We consider the estimation of a random vector observed through a linear transform followed by a componentwise probabilistic measurement channel. Although such linear mixing estimation problems are generally highly non-convex, Gaussian approximations of belief propagation (BP) have proven to be computationally attractive and highly effective in a range of applications. Recently, Bayati and Montanari have provided a rigorous and extremely general analysis of a large class of approximate message pa...
428 CitationsSource
#1David L. Donoho (Stanford University)H-Index: 99
#2Jared Tanner (Edin.: University of Edinburgh)H-Index: 24
We review connections between phase transitions in high-dimensional combinatorial geometry and phase transitions occurring in modern high-dimensional data analysis and signal processing. In data analysis, such transitions arise as abrupt breakdown of linear model selection, robust data fitting or compressed sensing reconstructions, when the complexity of the model or the number of outliers increases beyond a threshold. In combinatorial geometry, these transitions appear as abrupt changes in the ...
286 CitationsSource
#1David L. Donoho (Stanford University)H-Index: 99
#2Arian Maleki (Stanford University)H-Index: 23
Last. Andrea Montanari (Stanford University)H-Index: 49
view all 3 authors...
Compressed sensing aims to undersample certain high-dimensional signals yet accurately reconstruct them by exploiting signal characteristics. Accurate reconstruction is possible when the object to be recovered is sufficiently sparse in a known basis. Currently, the best known sparsity–undersampling tradeoff is achieved when reconstructing by convex optimization, which is expensive in important large-scale applications. Fast iterative thresholding algorithms have been intensively studied as alter...
1,188 CitationsSource
Cited By3
#1Jing Hua (JXAU: Jiangxi Agricultural University)
#2Yilu Xu (JXAU: Jiangxi Agricultural University)
Last. Jihao Zhang (HUST: Huazhong University of Science and Technology)
view all 5 authors...
Abstract The heartbeat classification of ECG signals on wearable devices has attracted extensive attention in recent years. Many existing works have studied them, but they do not consider the energy consumption of wearable device for classification of ECG signals. These methods are thus not suitable for wearable devices. In this paper, we propose a novel ECG heartbeat classification scheme performed in the compressive domain to reduce energy consumption for wearable devices. Specifically, we dev...
#1Henry Braun (ASU: Arizona State University)H-Index: 6
#2Pavan Turaga (ASU: Arizona State University)H-Index: 24
Last. Cihan Tepedelenlioglu (ASU: Arizona State University)H-Index: 26
view all 6 authors...
Abstract Compressed sensing (CS) allows signals and images to be reliably inferred from undersampled measurements. Exploiting CS allows the creation of new types of high-performance sensors includi...
#1Andreas Spanias (ASU: Arizona State University)H-Index: 28
Photovoltaic (PV) array analytics and control have become necessary for remote solar farms and for intelligent fault detection and power optimization. The management of a PV array requires auxiliary electronics that are attached to each solar panel. A collaborative industry-university-government project was established to create a smart monitoring device (SMD) and establish associated algorithms and software for fault detection and solar array management. First generation smart monitoring device...
11 CitationsSource
#1Sunil Rao (ASU: Arizona State University)H-Index: 4
#2Sameeksha Katoch (ASU: Arizona State University)H-Index: 3
Last. Devarajan SrinivasanH-Index: 5
view all 11 authors...
In this paper, we describe a Cyber-Physical system approach to Photovoltaic (PV) array control. A machine learning and computer vision framework is proposed for improving the reliability of utility scale PV arrays by leveraging video analysis of local skyline imagery, customized machine learning methods for fault detection, and monitoring devices that sense data and actuate at each individual panel. Our approach promises to improve efficiency in renewable energy systems using cyber-enabled senso...
10 CitationsSource
Dec 1, 2016 in ISSPIT (International Symposium on Signal Processing and Information Technology)
#1Jongmin Lee (ASU: Arizona State University)H-Index: 4
#2Michael Stanley (NXP Semiconductors)H-Index: 2
Last. Cihan Tepedelenlioglu (ASU: Arizona State University)H-Index: 26
view all 4 authors...
Interpreting sensor data in Internet-of-Things applications is a challenging problem particularly in embedded systems. We consider sensor data analytics where machine learning algorithms can be fully implemented on an embedded processor/sensor board. We develop an efficient real-time realization of a Gaussian mixture model (GMM) for execution on the NXP FRDM-K64F embedded sensor board. We demonstrate the design of a customized program and data structure that generates real-time sensor features, ...
9 CitationsSource