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

Proposes a new variable step LMS algorithm for transform domain (TDVSLMS). The new algorithm has the advantage of a variable step-size and the decorrelation of the input signal for increased speed of convergence. The difference between the new algorithm and the existing transform domain approaches is that the update of the step-size is based on the output error. In the case of the new TDVSLMS algorithm the step-size for each filter tap contains two time-variable components: the local component given by the normalization with the coefficient power, and the overall component that depends on the output error. In the simulations presented we show the benefits of this new algorithm compared with other time domain and transform domain implementations.

A new variable step LMS algorithm for transform domain

We discuss the performance of a set of local transform domain filters, with respect to varying parameters. We perform Monte Carlo simulations of the filters over five different test images at different noise levels, and evaluate how much the selection of filter parameters affect the filter performance and how far their performances are from that of the ideal Wiener filter.

Evaluation of potentials for local transform domain filters with varying parameters

RGB-D sensors are relatively inexpensive and are commercially available off-the-shelf. However, owing to their low complexity, there are several artifacts that one encounters in the depth map like holes, mis-alignment between the depth and color image and lack of sharp object boundaries in the depth map. Depth map generated by Kinect cameras also contain a significant amount of missing pixels and strong noise, limiting their usability in many computer vision applications. In this paper, we present an efficient hole filling and damaged region restoration method that improves the quality of the depth maps obtained with the Microsoft Kinect device. The proposed approach is based on a modified exemplar-based inpainting and LPA-ICI filtering by exploiting the correlation between color and depth values in local image neighborhoods. As a result, edges of the objects are sharpened and aligned with the objects in the color image. Several examples considered in this paper show the effectiveness of the proposed approach for large holes removal as well as recovery of small regions on several test images of depth maps. We perform a comparative study and show that statistically, the proposed algorithm delivers superior quality results compared to existing algorithms.

Depth map occlusion filling and scene reconstruction using modified exemplar-based inpainting

In this paper, an adaptive step-size LMS algorithm for tracking time-varying channels is presented. It is well known that in the case of such channels, the output steady-state mean square error (MSB) is a nonlinear function of the algorithm step-size and so, an optimum step-size that minimize the MSE exist. Here we propose an algorithm which adaptively adjust the step-size of the LMS toward its optimum value, such that the steady-state MSE is minimized. The nonlinear relation between the steady-state MSE and the step-size is parametrized such that, during the adaptation, estimates of the optimum step-size are easily obtained. These estimates are obtained independent from the channel statistical parameters, therefore, no prior information about the channel parameters is needed.

LMS adaptive filter with optimum step-size for tracking time-varying channels

Application of software complex for query processing in the database management system with novel noise resistant bispectrum-organized modulation technique in digital wireless communication systems

Karen Egiazarian

Papers

A new variable step LMS algorithm for transform domain

Evaluation of potentials for local transform domain filters with varying parameters

Depth map occlusion filling and scene reconstruction using modified exemplar-based inpainting

LMS adaptive filter with optimum step-size for tracking time-varying channels

Application of software complex for query processing in the database management system with novel noise resistant bispectrum-organized modulation technique in digital wireless communication systems