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

In this paper we consider an important practical aspect of texture region detection in remote sensing images. One specific feature of our study is that we assume a processed image noisy with a priori known type and parameters of the noise. Another specific feature is that we try to detect textural regions for a wide variety of textures without having a priori knowledge of their properties. The considered task is solved by means of trained neural networks. In the paper, we analyze the aspects of choosing input local parameters used in detection (recognition) and carrying out training. The verification results provide valuable conclusions for these aspects.

http://ieeexplore.ieee.org/iel7/7529352/7537982/07538174.pdf

Texture region detection by trained neural network

A novel phase retrieval algorithm for broadband hyperspectral phase imaging from noisy intensity observations is proposed. It utilizes advantages of the Fourier Transform spectroscopy in the self-referencing optical setup and provides, additionally beyond spectral intensity distribution, reconstruction of the investigated object's phase. The noise amplification Fellgett's disadvantage is relaxed by the application of sparse wavefront noise filtering embedded in the proposed algorithm. The algorithm reliability is proved by simulation tests and results of physical experiments on transparent objects which demonstrate precise phase imaging and object depth (profile) reconstructions.

Lensless hyperspectral imaging by Fourier transform spectroscopy for broadband visible light: phase retrieval technique

In this paper, we investigate a problem in computation of running discrete orthogonal transforms (RDOT). Due to overlapping blocks of signals, the RDOT for the signal block j can be computed recurrently by representing it as sum of tow terms. The first one is obtained by a multiplication of the circular advance matrix (CAM) by the RDOT of signal block j-1, and the second term is the transform of the sparse vector, formed from the weighted differences of the samples discarded from the block j-1 and the incoming samples of the block j. The second term of this RDOT decomposition could be efficiently implemented using fast prunned algorithms. The computational complexity of the RDOT depend mainly on the implementation of the first term. General conditions on the transform matrix for which CAM is completely sparse is established.© (1998) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Analysis of running discrete orthogonal transforms

Blind Multiframe Image Deconvolution Using Anisotropic Spatially Adaptive Filtering for Denoising and Regularization

Content–based image retrieval systems have plenty of applications in modern world. The most important one is the image search by query image or by semantic description. Approaches to this problem are employed in personal photo–collection management systems, web–scale image search engines, medical systems, etc. Automatic analysis of large unlabeled image datasets is virtually impossible without satisfactory image–retrieval technique. It’s the main reason why this kind of automatic image processing has attracted so much attention during recent years. Despite rather huge progress in the field, semantically meaningful image retrieval still remains a challenging task. The main issue here is the demand to provide reliable results in short amount of time. This paper addresses the problem by novel technique for simultaneous learning of global image features and binary hash codes. Our approach provide mapping of pixel–based image representation to hash–value space simultaneously trying to save as much of semantic image content as possible. We use deep learning methodology to generate image description with properties of similarity preservation and statistical independence. The main advantage of our approach in contrast to existing is ability to fine–tune retrieval procedure for very specific application which allow us to provide better results in comparison to general techniques. Presented in the paper framework for data– dependent image hashing is based on use two different kinds of neural networks: convolutional neural networks for image description and autoencoder for feature to hash space mapping. Experimental results confirmed that our approach has shown promising results in compare to other state–of–the–art methods.

Karen Egiazarian

Papers

Texture region detection by trained neural network

Lensless hyperspectral imaging by Fourier transform spectroscopy for broadband visible light: phase retrieval technique

Analysis of running discrete orthogonal transforms

Blind Multiframe Image Deconvolution Using Anisotropic Spatially Adaptive Filtering for Denoising and Regularization

Simultenious binary hash and features learning for image retrieval