Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Data Mining Practical Machine Learning Tools and Techniques

A survey on deep learning in medical image analysis

CA : A Cancer Journal for Clinicians

In the last years, wireless sensor networks (WSNs) have gained increasing attention from both the research community and actual users. As sensor nodes are generally battery-powered devices, the critical aspects to face concern how to reduce the energy consumption of nodes, so that the network lifetime can be extended to reasonable times. In this paper we first break down the energy consumption for the components of a typical sensor node, and discuss the main directions to energy conservation in WSNs. Then, we present a systematic and comprehensive taxonomy of the energy conservation schemes, which are subsequently discussed in depth. Special attention has been devoted to promising solutions which have not yet obtained a wide attention in the literature, such as techniques for energy efficient data acquisition. Finally we conclude the paper with insights for research directions about energy conservation in WSNs.

/pdf/energy-conservation-in-wireless-sensor-networks-a-survey-1bl7d2obnm.pdf

Energy conservation in wireless sensor networks: A survey

In recent years, sensor research has been undergoing a quiet revolution, promising to have a significant impact throughout society that could quite possibly dwarf previous milestones in the information revolution. Realizing the great promise of sensor networks requires more than a mere advance in individual technologies. It relies on many components working together in an efficient, unattended, comprehensible, and trustworthy manner. One of the enabling technologies in sensor networks is the distributed source coding (DSC), which refers to the compression of the multiple correlated sensor outputs that does not communicate with each other. DSC allows a many-to-one video coding paradigm that effectively swaps encoder-decoder complexity with respect to conventional video coding, thereby representing a fundamental concept shift in video processing. This article has presented an intensive discussion on two DSC techniques, namely Slepian-Wolf coding and Wyner-Ziv coding. The Slepian and Wolf coding have theoretically shown that separate encoding is as efficient as joint coding for lossless compression in channel coding.

Distributed source coding for sensor networks

We study lossy-to-lossless compression of medical volumetric data using three-dimensional (3-D) integer wavelet transforms. To achieve good lossy coding performance, it is important to have transforms that are unitary. In addition to the lifting approach, we first introduce a general 3-D integer wavelet packet transform structure that allows implicit bit shifting of wavelet coefficients to approximate a 3-D unitary transformation. We then focus on context modeling for efficient arithmetic coding of wavelet coefficients. Two state-of-the-art 3-D wavelet video coding techniques, namely, 3-D set partitioning in hierarchical trees (Kim et al., 2000) and 3-D embedded subband coding with optimal truncation (Xu et al., 2001), are modified and applied to compression of medical volumetric data, achieving the best performance published so far in the literature-both in terms of lossy and lossless compression.

Lossy-to-lossless compression of medical volumetric data using three-dimensional integer wavelet transforms

Compressive sampling is a novel framework that exploits sparsity of a signal in a transform domain to perform sampling below the Nyquist rate. In this paper, we apply compressive sampling to significantly reduce the sampling rate of video. A practical system is developed that first splits each video frame into non-overlapping blocks of equal size. Compressive sampling is then performed on sparse blocks, determined by predicting sparsity based on previous reference frames which are sampled conventionally. The blocks identified as sparse are reconstructed using the orthogonal matching pursuit algorithm, whereas the remaining blocks are sampled fully. Thus, the acquisition complexity and sampling time are reduced, while exploiting the local sparsity, within the DCT domain, of a video stream. Our simulation results indicate up to 50% saving in acquisition for Y-components of video with very small performance loss compared to traditional sampling.

/pdf/compressive-sampling-of-binary-images-3bg6x690rs.pdf

Compressive Sampling of Binary Images

This paper considers trellis coded quantization (TCQ) and low-density parity-check (LDPC) codes for the quadratic Gaussian Wyner-Ziv coding problem. After TCQ of the source X, LDPC codes are used to implement Slepian-Wolf coding of the quantized source Q(X) with side information Y at the decoder. Assuming 256-state TCQ and ideal Slepian-Wolf coding in the sense of achieving the theoretical limit H(Q(X)|Y ), we experimentally show that Slepian-Wolf coded TCQ performs 0.2 dB away from the Wyner-Ziv distortion-rate function DWZ(R) at high rate. This result mirrors that of entropy-constrained TCQ in classic source coding of Gaussian sources. Furthermore, using 8,192-state TCQ and assuming ideal Slepian-Wolf coding, our simulations show that Slepian-Wolf coded TCQ performs only 0.1 dB away from DWZ(R) at high rate. These results establish the practical performance limit of Slepian-Wolf coded TCQ for quadratic Gaussian Wyner-Ziv coding. Practical designs give performance very close to the theoretical limit. For example, with 8,192-state TCQ, irregular LDPC codes for Slepian-Wolf coding and optimal non-linear estimation at the decoder, our performance gap to DWZ(R) is 0.20 dB, 0.22 dB, 0.30 dB, and 0.93 dB at 3.83 bit per sample (b/s), 1.83 b/s, 1.53 b/s, and 1.05 b/s, respectively. When 256-state 4-D trellis-coded vector quantization instead of TCQ is employed, the performance gap to DWZ(R) is 0.51 dB, 0.51 dB, 0.54 dB, and 0.80 dB at 2.04 b/s, 1.38 b/s, 1.0 b/s, and 0.5 b/s, respectively.

/pdf/wyner-ziv-coding-based-on-tcq-and-ldpc-codes-1xjy0agmqo.pdf

Wyner-Ziv coding based on TCQ and LDPC codes

It has recently been demonstrated that two-dimensional photonic band gaps can be realized in systems comprising of a quasiperiodic arrangement of dielectric cylinders. We show that waveguides crafted out of such photonic ``quasicrystals'' can be used to guide light around sharp corners, just as in the case of periodic photonic band-gap systems, but the quasiperiodic systems tend to be more frequency selective. Because of the absence of translational symmetry, these quasiperiodic photonic band-gap structures also display richer defect properties. Spectral gaps for both the TM and TE polarizations in the same frequency range can be realized in metallodielectric configurations.

/pdf/defect-and-transmission-properties-of-two-dimensional-59pe3lfsrh.pdf

Samuel Cheng

Papers

Distributed source coding for sensor networks

Lossy-to-lossless compression of medical volumetric data using three-dimensional integer wavelet transforms

Compressive Sampling of Binary Images

Wyner-Ziv coding based on TCQ and LDPC codes

Defect and transmission properties of two-dimensional quasiperiodic photonic band-gap systems