An overview of the self-organizing map algorithm, on which the papers in this issue are based, is presented in this article.

The Self-Organizing Map

Convolutional neural networks CNNs have been applied to visual tasks since the late 1980s. However, despite a few scattered applications, they were dormant until the mid-2000s when developments in computing power and the advent of large amounts of labeled data, supplemented by improved algorithms, contributed to their advancement and brought them to the forefront of a neural network renaissance that has seen rapid progression since 2012. In this review, which focuses on the application of CNNs to image classification tasks, we cover their development, from their predecessors up to recent state-of-the-art deep learning systems. Along the way, we analyze 1 their early successes, 2 their role in the deep learning renaissance, 3 selected symbolic works that have contributed to their recent popularity, and 4 several improvement attempts by reviewing contributions and challenges of over 300 publications. We also introduce some of their current trends and remaining challenges.

Deep convolutional neural networks for image classification: A comprehensive review

In recent years, neural network accelerators have been shown to achieve both high energy efficiency and high performance for a broad application scope within the important category of recognition and mining applications. Still, both the energy efficiency and performance of such accelerators remain limited by memory accesses. In this paper, we focus on image applications, arguably the most important category among recognition and mining applications. The neural networks which are state-of-the-art for these applications are Convolutional Neural Networks (CNN), and they have an important property: weights are shared among many neurons, considerably reducing the neural network memory footprint. This property allows to entirely map a CNN within an SRAM, eliminating all DRAM accesses for weights. By further hoisting this accelerator next to the image sensor, it is possible to eliminate all remaining DRAM accesses, i.e., for inputs and outputs. In this paper, we propose such a CNN accelerator, placed next to a CMOS or CCD sensor. The absence of DRAM accesses combined with a careful exploitation of the specific data access patterns within CNNs allows us to design an accelerator which is 60&times more energy efficient than the previous state-of-the-art neural network accelerator. We present a full design down to the layout at 65 nm, with a modest footprint of 4.86mm2 and consuming only 320mW, but still about 30× faster than high-end GPUs.

ShiDianNao: shifting vision processing closer to the sensor

Theory and applications of HVAC control systems – A review of model predictive control (MPC)

We present a novel method for real-time continuous pose recovery of markerless complex articulable objects from a single depth image. Our method consists of the following stages: a randomized decision forest classifier for image segmentation, a robust method for labeled dataset generation, a convolutional network for dense feature extraction, and finally an inverse kinematics stage for stable real-time pose recovery. As one possible application of this pipeline, we show state-of-the-art results for real-time puppeteering of a skinned hand-model.

Real-Time Continuous Pose Recovery of Human Hands Using Convolutional Networks

Automatic recognition of gestures using computer vision is important for many real-world applications such as sign language recognition and human-robot interaction (HRI). Our goal is a real-time hand gesture-based HRI interface for mobile robots. We use a state-of-the-art big and deep neural network (NN) combining convolution and max-pooling (MPCNN) for supervised feature learning and classification of hand gestures given by humans to mobile robots using colored gloves. The hand contour is retrieved by color segmentation, then smoothened by morphological image processing which eliminates noisy edges. Our big and deep MPCNN classifies 6 gesture classes with 96% accuracy, nearly three times better than the nearest competitor. Experiments with mobile robots using an ARM 11 533MHz processor achieve real-time gesture recognition performance.

/pdf/max-pooling-convolutional-neural-networks-for-vision-based-3adx83uq6g.pdf

Max-pooling convolutional neural networks for vision-based hand gesture recognition

This letter extends previous research work in modeling a nontechnical loss (NTL) framework for the detection of fraud and electricity theft in power distribution utilities. Previous work was carried out by using a support vector machine (SVM)-based NTL detection framework resulting in a detection hitrate of 60%. This letter presents the inclusion of human knowledge and expertise into the SVM-based fraud detection model (FDM) with the introduction of a fuzzy inference system (FIS), in the form of fuzzy IF-THEN rules. The FIS acts as a postprocessing scheme for short-listing customer suspects with higher probabilities of fraud activities. With the implementation of this improved SVM-FIS computational intelligence FDM, Tenaga Nasional Berhad Distribution's detection hitrate has increased from 60% to 72%, thus proving to be cost effective.

Improving SVM-Based Nontechnical Loss Detection in Power Utility Using the Fuzzy Inference System

Mammography is currently the most effective imaging modality used by radiologists for the screening of breast cancer Finding an accurate, robust and efficient breast profile segmentation technique still remains a challenging problem in digital mammography Extraction of the breast profile region and the pectoral muscle is an essential pre-processing step in the process of computer-aided detection Primarily it allows the search for abnormalities to be limited to the region of the breast tissue without undue influence from the background of the mammogram The presence of pectoral muscle in mammograms biases detection procedures, which recommends removing the pectoral muscle during mammogram pre-processing In this paper we explore an automated technique for mammogram segmentation The proposed algorithm uses morphological preprocessing and seeded region growing (SRG) algorithm in order to: (1) remove digitization noises, (2) suppress radiopaque artifacts, (3) separate background region from the breast profile region, and (4) remove the pectoral muscle, for accentuating the breast profile region To demonstrate the capability of our proposed approach, digital mammograms from two separate sources are tested using Ground Truth (GT) images for evaluation of performance characteristics Experimental results obtained indicate that the breast regions extracted accurately correspond to the respective GT images

Automated breast profile segmentation for ROI detection using digital mammograms

Forecasting of future electricity demand is very important for decision making in power system operation and planning. In recent years, due to privatization and deregulation of the power industry, accurate electricity forecasting has become an important research area for efficient electricity production. This paper presents a time series approach for mid-term load forecasting (MTLF) in order to predict the daily peak load for the next month. The proposed method employs a computational intelligence scheme based on the self-organizing map (SOM) and support vector machine (SVM). According to the similarity degree of the time series load data, SOM is used as a clustering tool to cluster the training data into two subsets, using the Kohonen rule. As a novel machine learning technique, the support vector regression (SVR) is used to fit the testing data based on the clustered subsets, for predicting the daily peak load. Our proposed SOM-SVR load forecasting model is evaluated in MATLAB on the electricity load dataset provided by the Eastern Slovakian Electricity Corporation, which was used in the 2001 European Network on Intelligent Technologies (EUNITE) load forecasting competition. Power load data obtained from (i) Tenaga Nasional Berhad (TNB) for peninsular Malaysia and (ii) PJM for the eastern interconnection grid of the United States of America is used to benchmark the performance of our proposed model. Experimental results obtained indicate that our proposed SOM-SVR technique gives significantly good prediction accuracy for MTLF compared to previously researched findings using the EUNITE, Malaysian and PJM electricity load datasets.

A computational intelligence scheme for the prediction of the daily peak load

Photovoltaic water pumping system (PVPS) is an important and promising application of solar energy systems especially in remote areas. In this review paper, research work on PVPS modeling, reliability, feasibility, field performance, design procedures and control strategies is analyzed and reported. In addition, technical challenges to this science are concluded in this review. According to the results, the numerical method is recommended for sizing PVPS. In the meanwhile, incremental conductance and artificial intelligent (AI) based maximum power point tracking algorithms are the most recommended control strategies for PVPS. On the other hand, According to reported PVPS performance, the average overall system's efficiency is about 3.4% with average annual productivity rate of 6580.611 m3/kWp. In general PVPS have a number of challenges related to design procedure, modeling method, control strategy, availability of data and site obstacles such as shadow effects.

Farrukh Nagi

Papers

Max-pooling convolutional neural networks for vision-based hand gesture recognition

Improving SVM-Based Nontechnical Loss Detection in Power Utility Using the Fuzzy Inference System

Automated breast profile segmentation for ROI detection using digital mammograms

A computational intelligence scheme for the prediction of the daily peak load

A review of photovoltaic water pumping system designing methods, control strategies and field performance