We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

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

计量经济分析 = Econometric analysis

This paper introduces novel binary and multilevel phase masks (BPMs) for improved depth of focus (DoF) in infrared imaging. The procedure developed for design of BPMs combines two ideas: cubic wavefront coding (WFC) of continuous absolute phase and original discretization of this phase profile. Both these ingredients of the design make the optical system robust with respect to defocus and uncontrolled variations in the optical system. The method allows us to design a flat and very thin BPM using piecewise invariant diffractive optical element (DOE) of simple geometry. The BPMs are used for wavefront modulation in two optical setups: a lensless system and a lens/BPM optical hybrid. Computational inverse imaging (deblurring) is applied in order to reconstruct a sharp image from respective blurred observations. The optical system is optimized end-to-end, including both the BPMs and the deblurring algorithm. Simulation experiments demonstrated for midwave infrared (MWIR) imaging, show high quality imaging even for very large amounts of defocus. Imaging systems incorporating this method show advantages over conventional optical systems with a refractive lens.

A novel binary and multilevel phase masks for enhanced depth-of-focus infrared imaging

http://www.dl.begellhouse.com/download/article/4179fa176ddc1ebf/139-154.pdf

On requirements to accuracy of noise variance estimation in prediction of dct-based filter efficiency

The paper is devoted to the task of estimation of the parameters of spatially correlated noise and noise suppression in images. Several schemes of noise removal, including multiscale ones, are considered. A convolutional neural network (CNN) for blind estimation of the spectrum of spatially correlated noise images is proposed. It is shown that the proposed network in combination with the BM3D filter provides more efficient noise suppression than existing solutions. A CNN for prediction of the denoising parameters for DRUNet denoiser is also proposed and analyzed. It is shown that the usage of this network and DRUNet for multiscale denoising in comparison with other methods provides better quality of image denoising and processing speed for a wide range of sizes of “noise grain”.

Blind Estimation and Suppression of Additive Spatially Correlated Gaussian Noise in Images

Two methods of the orthogonal spectra extrapolation problem are considered in this paper. Both solutions are based on the application of multi-valued neural element as a filter and as an extrapolator. The first method consist of approximation of the spectra in the higher frequency part using an iterative approximation and multi-valued non-linear filtering. The second method is reduced to the prediction of the spectral coefficients corresponding to the higher frequency part using possibility of MVN to predict time-series. Application of the proposed methods to solution of the super-resolution problem are presented.

Solution of the super-resolution problem through extrapolation of the orthogonal spectra using multi-valued neural technique

The reconstruction of 1-D signal waveform from noisy bispectrum estimations in case of a limited ensemble realization number and low input SNRs leads to considerable output errors. In order to improve the signal bispectrum estimations we propose to smooth the obtained bispectrum estimations. The methods based on vector filter application to 2-D complex valued bispectrum processing are considered in this paper. The influence of input SNR, the filter type and the used norm selection on signal reconstruction system performance is analyzed using computer simulations. It is shown that the proposed methods decreases both fluctuation error and bias of reconstructed signal.

Karen Egiazarian

Papers

A novel binary and multilevel phase masks for enhanced depth-of-focus infrared imaging

On requirements to accuracy of noise variance estimation in prediction of dct-based filter efficiency

Blind Estimation and Suppression of Additive Spatially Correlated Gaussian Noise in Images

Solution of the super-resolution problem through extrapolation of the orthogonal spectra using multi-valued neural technique

Signal waveform reconstruction from noisy bispectrum estimations pre-processed by vector filters