A Similarity-Based Object Classification Using Deep Neural Networks

This chapter proposes a hybrid method for classification of the objects based on deep neural network and a similarity-based search algorithm. The objects are pre-processed with external conditions. After pre-processing and training different deep learning networks with the object dataset, the authors compare the results to find the best model to improve the accuracy of the results based on the features of object images extracted from the feature vector layer of a neural network. RPFOREST (random projection forest) model is used to predict the approximate nearest images. ResNet50, InceptionV3, InceptionV4, and DenseNet169 models are trained with this dataset. A proposal for adaptive finetuning of the deep learning models by determining the number of layers required for finetuning with the help of the RPForest model is given, and this experiment is conducted using the Xception model.

Detection of Blood-Related Diseases Using Deep Neural Nets

Image processing offers medical diagnosis and it overcomes the shortcomings faced by traditional laboratory methods with the help of intelligent algorithms. It is also useful for remote quality control and consultations. As machine learning is stepping into biomedical engineering, there is a huge demand for devices which are intelligent and accurate enough to target the diseases. The platelet count in a blood sample can be done by extrapolating the number of platelets counted in the blood smear. Deep neural nets use multiple layers of filtering and automated feature extraction and detection and can overcome the hurdle of devising complex algorithms to extract features for each type of disease. So, this chapter deals with the usage of deep neural networks for the image classification and platelets count. The method of using deep neural nets has increased the accuracy of detecting the disease and greater efficiency compared to traditional image processing techniques. The method can be further expanded to other forms of diseases which can be detected through blood samples.

Deep Clustering

Clustering is an unsupervised technique used in various application, namely machine learning, image segmentation, social network analysis, health analytics, and financial analysis. It is a task of grouping similar objects together and dissimilar objects in different group. The quality of the cluster relies on two factors: distance metrics and data representation. Deep learning is a new field of machine learning research that has been introduced to move machine learning closer to artificial intelligence. Learning using deep network provides multiple layers of representation that helps to understand images, sound, and text. In this chapter, the need for deep network in clustering, various architecture, and algorithms for unsupervised learning is discussed.

Deep Learning in Early Detection of Alzheimer's

Advancement in technology has paved the way for the growth of big data. We are able to exploit this data to a great extent as the costs of collecting, storing, and analyzing a large volume of data have plummeted considerably. There is an exponential increase in the amount of health-related data being generated by smart devices. Requisite for proper mining of the data for knowledge discovery and therapeutic product development is very essential. The expanding field of big data analytics is playing a vital role in healthcare practices and research. A large number of people are being affected by Alzheimer's Disease (AD), and as a result, it becomes very challenging for the family members to handle these individuals. The objective of this chapter is to highlight how deep learning can be used for the early diagnosis of AD and present the outcomes of research studies of both neurologists and computer scientists. The chapter gives introduction to big data, deep learning, AD, biomarkers, and brain images and concludes by suggesting blood biomarker as an ideal solution for early detection of AD.

Enhanced Footsteps Generation Method for Walking Robots Based on Convolutional Neural Networks

In this chapter, the problem of finding a suitable foothold for a bipedal walking robot is studied. There are a number of gait generation algorithms that rely on having a set of obstacle-free regions where the robot can step to and there are a number of algorithms for generating these regions. This study breaches the gap between these algorithms, providing a way to quickly check if a given obstacle free region is accessible for foot placement. The proposed approach is based on the use of a classifier, constructed as a convolutional neural network. The study discusses the training dataset generation, including datasets with uncertainty related to the shapes of the obstacle-free regions. Training results for a number of different datasets and different hyperparameter choices are presented and showed robustness of the proposed network design both to different hyperparameter choices as well as to the changes in the training dataset.

Anomaly Detection Using Deep Learning With Modular Networks

In daily realistic activities, security is one of the main criteria among the different machines like IOT devices, networks. In these systems, anomaly detection is one of the issues. Anomaly detection based on user behavior is very essential to secure the machines from the unauthorized activities by anomaly user. Techniques used for an anomaly detection is to learn the daily realistic activities of the user, and later it proactively detects the anomalous situation and unusual activities. In the IOT-related systems, the detection of such anomalous situations can be fine-tuned with minor and major erroneous conditions to the machine learning algorithms that learn the activities of a user. In this chapter, neural networks, with multiple hidden layers to detect the different situation by creating an environment with random anomalous activities to the machine, are proposed. Using deep learning for anomaly detection would help in enhancing the accuracy and speed.

Deep Reinforcement Learning for Optimization

Deep reinforcement learning (DRL) has transformed the field of artificial intelligence (AI) especially after the success of Google DeepMind. This branch of machine learning epitomizes a step toward building autonomous systems by understanding of the visual world. Deep reinforcement learning (RL) is currently applied to different sorts of problems that were previously obstinate. In this chapter, at first, the authors started with an introduction of the general field of RL and Markov decision process (MDP). Then, they clarified the common DRL framework and the necessary components RL settings. Moreover, they analyzed the stochastic gradient descent (SGD)-based optimizers such as ADAM and a non-specific multi-policy selection mechanism in a multi-objective Markov decision process. In this chapter, the authors also included the comparison for different Deep Q networks. In conclusion, they describe several challenges and trends in research within the deep reinforcement learning field.

A Survey on Deep Learning Techniques Used for Quality Process

In the current era, software development and software quality has become extensively important for implementing the real-world software application, and it will enhance the software functionality. Moreover, early prediction of expected error and fault level in the quality process is critical to the software development process. Deep learning techniques are the most appropriate methods for this problem, and this chapter carries out an extensive systematic survey on a variety of deep learning. These techniques are used in the software quality process along with a hypothesis justification for each of the proposed solutions. The deep learning and machine learning techniques are considered to be the most suitable systems for software quality prediction. Deep learning is a computational model made up of various hidden layers of investigation used to portray of information with the goal that researchers can better understand complex information issues.

CNN Customizations With Transfer Learning for Face Recognition Task

Several kinds of pretrained convolutional neural networks (CNN) exist nowadays. Utilizing these networks with the new classification task requires the retraining with new data sets. With the small embedded device, large network cannot be implemented. The authors study the use of pretrained models and customizing them towards accuracy and size against face recognition tasks. The results show 1) the performance of existing pretrained networks (e.g., AlexNet, GoogLeNet, CaffeNet, SqueezeNet), as well as size, and 2) demonstrate the layers customization towards the model size and accuracy. The studied results show that among the various networks with different data sets, SqueezeNet can achieve the same accuracy (0.99) as others with small size (up to 25 times smaller). Secondly, the two customizations with layer skipping are presented. The experiments show the example of SqueezeNet layer customizing, reducing the network size while keeping the accuracy (i.e., reducing the size by 7% with the slower convergence time). The experiments are measured based on Caffe 0.15.14.

Malware Classification and Analysis Using Convolutional and Recurrent Neural Network

Over the past decade, malware has grown exponentially. Traditional signature-based approaches to detecting malware have proven their limitations against new malware, and categorizing malware samples has become essential to understanding the basics of malware behavior. Recently, antivirus solutions have increasingly started to adopt machine learning approaches. Unfortunately, there are few open source data sets available for the academic community. One of the largest data sets available was published last year in a competition on Kaggle with data provided by Microsoft for the big data innovators gathering. This chapter explores the problem of malware classification. In particular, this chapter proposes an innovative and scalable approach using convolutional neural networks (CNN) and long short-term memory (LSTM) to assign malware to the corresponding family. The proposed method achieved a classification accuracy of 98.73% and an average log loss of 0.0698 on the validation data.