<div>Developing intelligent agents that can perceive and understand the rich visual world around us has been a long-standing goal in the field of AI. Recently, a significant progress has been made by the CNNs/ DNNs to the incredible advances & in a wide range of applications such as ADAS, intelligent cameras surveillance, autonomous systems, drones, & robots. Design space exploration (DSE) of NNs and other techniques have made CNN/ DNN memory & computationally efficient. But the major design hurdles for deployment are limited resources such as computation, memory, energy efficiency, and power budget. DSE of small DNN architectures for ADAS emerged with better and efficient architectures such as baseline SqueezeNet and SqueezeNext. These architectures are exclusively known for their small model size, good model speed & model accuracy. In this thesis study, two new DNN architectures are proposed. Before diving into the proposed architectures, DSE of DNNs explores the methods to improve DNNs/ CNNs. Further, understanding the different hyperparameters tuning & experimenting with various optimizers and newly introduced methodologies. First, High Performance SqueezeNext architecture ameliorate the performance of existing DNN architectures.The intuition behind this proposed architecture is to supplant convolution layers with a more sophisticated block module & to develop a compact and efficient architecture with a competitive accuracy. Second, Shallow SqueezeNext architecture is proposed which achieves better model size results in comparison to baseline SqueezeNet and SqueezeNext is presented. It illustrates the architecture is compact, efficient and flexible in terms of model size and accuracy. The state-of-the-art SqueezeNext baseline and SqueezeNext baseline are used as the foundation to recreate and propose the both DNN architectures in this study. Due to very small model size with competitive model accuracy and decent model testing speed it is expected to perform well on the ADAS systems. The proposed architectures are trained and tested from scratch on CIFAR-10 & CIFAR-100 datasets. All the training and testing results are visualized with live loss and accuracy graphs by using livelossplot. In the last, both of the proposed DNN architectures are deployed on BlueBox2.0 by NXP.</div>

Indiana University-Purdue University Indianapolis (IUPUI) / Developing intelligent agents that can perceive and understand the rich visualworld around us has been a long-standing goal in the field of AI. Recently, asignificant progress has been made by the CNNs/DNNs to the incredible advances& in a wide range of applications such as ADAS, intelligent cameras surveillance,autonomous systems, drones, & robots. Design space exploration (DSE) of NNs andother techniques have made CNN/DNN memory & computationally efficient. Butthe major design hurdles for deployment are limited resources such as computation,memory, energy efficiency, and power budget. DSE of small DNN architectures forADAS emerged with better and efficient architectures such as baseline SqueezeNetand SqueezeNext. These architectures are exclusively known for their small modelsize, good model speed & model accuracy.In this thesis study, two new DNN architectures are proposed. Before diving intothe proposed architectures, DSE of DNNs explores the methods to improveDNNs/CNNs.Further, understanding the different hyperparameters tuning &experimenting with various optimizers and newly introduced methodologies. First,High Performance SqueezeNext architecture ameliorate the performance of existingDNN architectures. The intuition behind this proposed architecture is to supplantconvolution layers with a more sophisticated block module & to develop a compactand efficient architecture with a competitive accuracy. Second, Shallow SqueezeNextarchitecture is proposed which achieves better model size results in comparison tobaseline SqueezeNet and SqueezeNext is presented. It illustrates the architecture is xviicompact, efficient and flexible in terms of model size and accuracy.Thestate-of-the-art SqueezeNext baseline and SqueezeNext baseline are used as thefoundation to recreate and propose the both DNN architectures in this study. Dueto very small model size with competitive model accuracy and decent model testingspeed it is expected to perform well on the ADAS systems.The proposedarchitectures are trained and tested from scratch on CIFAR-10 [30] & CIFAR-100[34] datasets. All the training and testing results are visualized with live loss andaccuracy graphs by using livelossplot. In the last, both of the proposed DNNarchitectures are deployed on BlueBox2.0 by NXP.

abstract: Machine learning (ML) and deep neural networks (DNNs) have achieved great success in a variety of application domains, however, despite significant effort to make these networks robust, they remain vulnerable to adversarial attacks in which input that is perceptually indistinguishable from natural data can be erroneously classified with high prediction confidence. Works on defending against adversarial examples can be broadly classified as correcting or detecting, which aim, respectively at negating the effects of the attack and correctly classifying the input, or detecting and rejecting the input as adversarial. In this work, a new approach for detecting adversarial examples is proposed. The approach takes advantage of the robustness of natural images to noise. As noise is added to a natural image, the prediction probability of its true class drops, but the drop is not sudden or precipitous. The same seems to not hold for adversarial examples. In other word, the stress response profile for natural images seems different from that of adversarial examples, which could be detected by their stress response profile. An evaluation of this approach for detecting adversarial examples is performed on the MNIST, CIFAR-10 and ImageNet datasets. Experimental data shows that this approach is effective at detecting some adversarial examples on small scaled simple content images and with little sacrifice on benign accuracy. / Dissertation/Thesis / Masters Thesis Computer Science 2019

<p dir="ltr">Humans exhibit remarkable ability in continual adaptation and learning new tasks throughout their lifetime while maintaining the knowledge gained from past experiences. In stark contrast, artificial neural networks (ANNs) under such continual learning (CL) paradigm forget the information learned in the past tasks upon learning new ones. This phenomenon is known as ‘Catastrophic Forgetting’ or ‘Catastrophic Interference’. The objective of this thesis is to enable efficient continual learning in deep neural networks while mitigating this forgetting phenomenon. Towards this, first, a continual learning algorithm (SPACE) is proposed where a subset of network filters or neurons is allocated for each task using Principal Component Analysis (PCA). Such task-specific network isolation not only ensures zero forgetting but also creates structured sparsity in the network which enables energy-efficient inference. Second, a fast and more efficient training algorithm for CL is proposed by introducing Gradient Projection Memory (GPM). Here, the most important gradient spaces (GPM) for each task are computed using Singular Value Decomposition (SVD) and the new tasks are learned in the orthogonal direction to GPM to minimize forgetting. Third, to improve new learning while minimizing forgetting, a Scaled Gradient Projection (SGP) method is proposed that, in addition to orthogonal gradient updates, allows scaled updates along the important gradient spaces of the past task. Next, for continual learning on an online stream of tasks a memory efficient experience replay method is proposed. This method utilizes saliency maps explaining network’s decision for selecting memories that are replayed during new tasks for preventing forgetting. Finally, a meta-learning based continual learner - Amphibian - is proposed that achieves fast online continual learning without any experience replay. All the algorithms are evaluated on short and long sequences of tasks from standard image-classification datasets. Overall, the methods proposed in this thesis address critical limitations of DNNs for continual learning and advance the state-of-the-art in this domain.</p>

<p>Deep Learning (DL)-based edge intelligence has garnered significant attention in recent years due to the rapid proliferation of the Internet of Things (IoT), embedded, and intelligent systems, collectively termed edge devices. Sensor data streams acquired by these edge devices are processed by a Deep Neural Network (DNN) application that runs on the device itself or in the cloud. However, the high computational complexity and energy consumption of processing DNNs often limit their deployment on these edge inference systems due to limited compute, memory and energy resources. Furthermore, high costs, strict application latency demands, data privacy, security constraints, and the absence of reliable edge-cloud network connectivity heavily impact edge application efficiency in the case of cloud-assisted DNN inference. Inevitably, performance and energy efficiency are of utmost importance in these edge inference systems, aside from the accuracy of the application. To facilitate energy- efficient edge inference systems running computationally complex DNNs, this dissertation makes three key contributions.</p> <p><br></p> <p>The first contribution adopts a full-system approach to Approximate Computing, a design paradigm that trades off a small degradation in application quality for significant energy savings. Within this context, we present the foundational concepts of AxIS, the first approximate edge inference system that jointly optimizes the constituent subsystems leading to substantial energy benefits compared to optimization of the individual subsystem. To illustrate the efficacy of this approach, we demonstrate multiple versions of an approximate smart camera system that executes various DNN-based unimodal computer vision applications, showcasing how the sensor, memory, compute, and communication subsystems can all be synergistically approximated for energy-efficient edge inference.</p> <p><br></p> <p>Building on this foundation, the second contribution extends AxIS to multimodal AI, harnessing data from multiple sensor modalities to impart human-like cognitive and perceptual abilities to edge devices. By exploring optimization techniques for multiple sensor modalities and subsystems, this research reveals the impact of synergistic modality-aware optimizations on system-level accuracy-efficiency (AE) trade-offs, culminating in the introduction of SysteMMX, the first AE scalable cognitive system that allows efficient multimodal inference at the edge. To illustrate the practicality and effectiveness of this approach, we present an in-depth case study centered around a multimodal system that leverages RGB and Depth sensor modalities for image segmentation tasks.</p> <p><br></p> <p>The final contribution focuses on optimizing the performance of an edge-cloud collaborative inference system through intelligent DNN partitioning and computation offloading. We delve into the realm of distributed inference across edge devices and cloud servers, unveiling the challenges associated with finding the optimal partitioning point in DNNs for significant inference latency speedup. To address these challenges, we introduce PArtNNer, a platform-agnostic and adaptive DNN partitioning framework capable of dynamically adapting to changes in communication bandwidth and cloud server load. Unlike existing approaches, PArtNNer does not require pre-characterization of underlying edge computing platforms, making it a versatile and efficient solution for real-world edge-cloud scenarios.</p> <p><br></p> <p>Overall, this thesis provides novel insights, innovative techniques, and intelligent solutions to enable energy-efficient AI at the edge. The contributions presented herein serve as a solid foundation for future researchers to build upon, driving innovation and shaping the trajectory of research in edge AI.</p>