This paper presents a study on the applicability of using approximate multipliers to enhance the performance of VGGNet deep learning network. Approximate multipliers are known to have reduced power, area, and delay with the cost of an inaccuracy in output. Improving the performance of the VGGNet in terms of power, area, and speed can be achieved by replacing exact multipliers with approximate multipliers as demonstrated in this study. The simulation results show that approximate multiplication has a very little impact on the accuracy of VGGNet. However, using approximate multipliers can achieve significant performance gains. The simulation was completed using different generated error matrices that mimic the inaccuracy that approximate multipliers introduce to the data. The impact of various ranges of the mean relative error (MRE) and the standard deviation (SD) was tested. The well-known datasets CIFAR-10 and CIFAR-100 were used for testing the network’s classification accuracy. The impact on the accuracy was assessed by simulating approximate multiplication in all the layers in the first set of tests, and in selective layers in the second set of tests. Using approximate multipliers in all the layers leads to very little impact on the network’s accuracy. Additionally, an alternative approach is to use a hybrid of exact and approximate multipliers. In the hybrid approach, 39.14% of the deeper layer’s multiplications can be approximate while having a reduced negligible impact on the network’s accuracy.

This article proposes boosting the multiplication performance for convolutional neural network (CNN) inference using a precision prediction preprocessor which controls various precision approximate multipliers. Previously, utilizing approximate multipliers for CNN inference was proposed to enhance the power, speed, and area at a cost of a tolerable drop in the accuracy. Low precision approximate multipliers can achieve massive performance gains; however, utilizing them is not feasible due to the large accuracy loss they cause. To maximize the multiplication performance gains while minimizing the accuracy loss, this article proposes using a tiny two-class precision controller to utilize low and high precision approximate multipliers hybridly. The performance benefits for the proposed concept are presented for multi-core multi-precision architectures and single-core reconfigurable architectures. Additionally, a design for a merged reconfigurable approximate multiplier with two precisions is proposed for utilization in single-core architectures. For performance comparison, several segments-based approximate multipliers with different precisions were synthesized using CMOS 15nm technology. For accuracy evaluation, the concept was simulated on VGG19, Xception, and DenseNet201 using the ImageNetV2 dataset. This article will demonstrate that the proposed concept can achieve significant performance gains with a minimal accuracy loss when compared to designs that utilize exact multipliers or single-precision approximate multipliers.

Accuracy evaluation in machine learning is based on the split of data into a training set and a test set. This critical step is applied to develop machine learning models including models based on sensor data. For sensor-based problems, comparing the accuracy of machine learning models using the train/test split provides only a baseline comparison in ideal situations. Such comparisons won’t consider practical production problems that can impact the inference accuracy such as the sensors’ thermal noise, performance with lower inference quantization, and tolerance to sensor failure. Therefore, this paper proposes a set of practical tests that can be applied when comparing the accuracy of machine learning models for sensor-based problems. First, the impact of the sensors’ thermal noise on the models’ inference accuracy was simulated. Machine learning algorithms have different levels of error resilience to thermal noise, as will be presented. Second, the models’ accuracy using lower inference quantization was compared. Lowering inference quantization leads to lowering the analog-to-digital converter (ADC) resolution which is cost-effective in embedded designs. Moreover, in custom designs, analog-to-digital converters’ (ADCs) effective number of bits (ENOB) is usually lower than the ideal number of bits due to various design factors. Therefore, it is practical to compare models’ accuracy using lower inference quantization. Third, the models’ accuracy tolerance to sensor failure was evaluated and compared. For this study, University of California Irvine (UCI) ‘Daily and Sports Activities’ dataset was used to present these practical tests and their impact on model selection.

A comparison of the performance of various machine learning models to predict the direction of a wall following robot is presented in this paper. The models were trained using an open-source dataset that contains 24 ultrasound sensors readings and the corresponding direction for each sample. This dataset was captured using SCITOS G5 mobile robot by placing the sensors on the robot waist. In addition to the full format with 24 sensors per record, the dataset has two simplified formats with 4 and 2 input sensor readings per record. Several control models were proposed previously for this dataset using all three dataset formats. In this paper, two primary research contributions are presented. First, presenting machine learning models with accuracies higher than all previously proposed models for this dataset using all three formats. A perfect solution for the 4 and 2 inputs sensors formats is presented using Decision Tree Classifier by achieving a mean accuracy of 100%. On the other hand, a mean accuracy of 99.82% was achieves using the 24 sensor inputs by employing the Gradient Boost Classifier. Second, presenting a comparative study on the performance of different machine learning and deep learning algorithms on this dataset. Therefore, providing an overall insight on the performance of these algorithms for similar sensor fusion problems. All the models in this paper were evaluated using Monte-Carlo cross-validation.

Nuclear reactor inspections are critical to ensure the safety and reliability of a nuclear facility’s operation. In Canada, ultrasonic testing (UT) is used to inspect the health of pressure tubes that are part of Canada’s Deuterium Uranium (CANDU) reactor’s fuel channels. Currently, analysis of UT scans is performed by manual visualization and measurement to locate, characterize, and disposition flaws. Therefore, there is motivation to develop an automated method that is fast and accurate. In this article, a proof of concept (PoC) that automates the detection of flaws in nuclear fuel channel UT scans using a convolutional neural network (CNN) is presented. The CNN model was trained after constructing a dataset using historical UT scans and the corresponding inspection results. The requirement for this prototype was to identify the location of at least a portion of each flaw in UT scans while minimizing false positives (FPs). The proposed CNN model achieves this target by automatically identifying at least a portion of each flaw where further manual analysis is performed to identify the width, the length, and the type of the flaw.

This article proposes EdgeCrypt Tracker, an object-tracking algorithm combined with differential encryption to provide better accuracy and runtime efficiency for battery-operated Internet of Autonomous Aerial Vehicles (IoAAV). Specifically, EdgeCrypt Tracker operates directly on high-efficiency video coding (HEVC) and has three stages: 1) preprocessing; 2) object tracking; and 3) differential encryption. The preprocessing stage separates intra frames and removes artificial camera motion caused by camera movement from inter frames. Next, the object tracking stage utilizes a hybrid neural network, combining a single-shot multibox detector (SSD) network with a MobileNetV3 backbone that processes intra coded blocks and a fast gated recurrent neural network (FastGRNN) network that processes inter coded blocks. Finally, the tracked information is passed to the differential encryption stage, which encrypts all syntax elements within moving objects and alternate syntax elements related to the background. Experimental results demonstrate that EdgeCrypt Tracker achieves an average object tracking accuracy of 92%, real-time inference with a 35% lower encryption overhead compared to state-of-the-art methods. This work demonstrates the potential of integrating object tracking and encryption within video compression for secure, efficient AAV-based surveillance.