Matching Items (8)
Filtering by
- All Subjects: deep learning
- All Subjects: Machine Learning
Description
In recent years, Artificial Intelligence (AI) (e.g., Deep Neural Networks (DNNs), Transformer) has shown great success in real-world applications due to its superior performance in various cognitive tasks. The impressive performance achieved by AI models normally accompanies the cost of enormous model size and high computational complexity, which significantly hampers their implementation on resource-limited Cyber-Physical Systems (CPS), Internet-of-Things (IoT), or Edge systems due to their tightly constrained energy, computing, size, and memory budget. Thus, the urgent demand for enhancing the \textbf{Efficiency} of DNN has drawn significant research interests across various communities.
Motivated by the aforementioned concerns, this doctoral research has been mainly focusing on Enabling Deep Learning at Edge: From Efficient and Dynamic Inference to On-Device Learning. Specifically, from the inference perspective, this dissertation begins by investigating a hardware-friendly model compression method that effectively reduces the size of AI model while simultaneously achieving improved speed on edge devices. Additionally, due to the fact that diverse resource constraints of different edge devices, this dissertation further explores dynamic inference, which allows for real-time tuning of inference model size, computation, and latency to accommodate the limitations of each edge device. Regarding efficient on-device learning, this dissertation starts by analyzing memory usage during transfer learning training. Based on this analysis, a novel framework called "Reprogramming Network'' (Rep-Net) is introduced that offers a fresh perspective on the on-device transfer learning problem. The Rep-Net enables on-device transferlearning by directly learning to reprogram the intermediate features of a pre-trained model. Lastly, this dissertation studies an efficient continual learning algorithm that facilitates learning multiple tasks without the risk of forgetting previously acquired knowledge. In practice, through the exploration of task correlation, an interesting phenomenon is observed that the intermediate features are highly correlated between tasks with the self-supervised pre-trained model. Building upon this observation, a novel approach called progressive task-correlated layer freezing is proposed to gradually freeze a subset of layers with the highest correlation ratios for each task leading to training efficiency.
ContributorsYang, Li (Author) / Fan, Deliang (Thesis advisor) / Seo, Jae-Sun (Committee member) / Zhang, Junshan (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2023
Description
Resistive random-access memory (RRAM) or memristor, is an emerging technology used in neuromorphic computing to exceed the traditional von Neumann obstacle by merging the processing and memory units. Two-dimensional (2D) materials with non-volatile switching behavior can be used as the switching layer of RRAMs, exhibiting superior behavior compared to conventional oxide-based RRAMs. The use of 2D materials allows scaling the resistive switching layer thickness to sub-nanometer dimensions enabling devices to operate with low switching voltages and high programming speeds, offering large improvements in efficiency and performance as well as ultra-dense integration. This dissertation presents an extensive study of linear and logistic regression algorithms implemented with 1-transistor-1-resistor (1T1R) memristor crossbars arrays. For this task, a simulation platform is used that wraps circuit-level simulations of 1T1R crossbars and physics-based model of RRAM to elucidate the impact of device variability on algorithm accuracy, convergence rate, and precision. Moreover, a smart pulsing strategy is proposed for the practical implementation of synaptic weight updates that can accelerate training in real crossbar architectures.
Next, this dissertation reports on the hardware implementation of analog dot-product operation on arrays of 2D hexagonal boron nitride (h-BN) memristors. This extends beyond previous work that studied isolated device characteristics towards the application of analog neural network accelerators based on 2D memristor arrays. The wafer-level fabrication of the memristor arrays is enabled by large-area transfer of CVD-grown few-layer h-BN films. The dot-product operation shows excellent linearity and repeatability, with low read energy consumption, with minimal error and deviation over various measurement cycles. Moreover, the successful implementation of a stochastic linear and logistic regression algorithm in 2D h-BN memristor hardware is presented for the classification of noisy images. Additionally, the electrical performance of novel 2D h-BN memristor for SNN applications is extensively investigated. Then, using the experimental behavior of the h-BN memristor as the artificial synapse, an unsupervised spiking neural network (SNN) is simulated for the image classification task. A novel and simple Spike-Timing-Dependent-Plasticity (STDP)-based dropout technique is presented to enhance the recognition task of the h-BN memristor-based SNN.
ContributorsAfshari, Sahra (Author) / Sanchez Esqueda, Ivan (Thesis advisor) / Barnaby, Hugh J (Committee member) / Seo, Jae-Sun (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2023
Description
Deep learning (DL) has proved itself be one of the most important developements till date with far reaching impacts in numerous fields like robotics, computer vision, surveillance, speech processing, machine translation, finance, etc. They are now widely used for countless applications because of their ability to generalize real world data, robustness to noise in previously unseen data and high inference accuracy. With the ability to learn useful features from raw sensor data, deep learning algorithms have out-performed tradinal AI algorithms and pushed the boundaries of what can be achieved with AI. In this work, we demonstrate the power of deep learning by developing a neural network to automatically detect cough instances from audio recorded in un-constrained environments. For this, 24 hours long recordings from 9 dierent patients is collected and carefully labeled by medical personel. A pre-processing algorithm is proposed to convert event based cough dataset to a more informative dataset with start and end of coughs and also introduce data augmentation for regularizing the training procedure. The proposed neural network achieves 92.3% leave-one-out accuracy on data captured in real world.
Deep neural networks are composed of multiple layers that are compute/memory intensive. This makes it difficult to execute these algorithms real-time with low power consumption using existing general purpose computers. In this work, we propose hardware accelerators for a traditional AI algorithm based on random forest trees and two representative deep convolutional neural networks (AlexNet and VGG). With the proposed acceleration techniques, ~ 30x performance improvement was achieved compared to CPU for random forest trees. For deep CNNS, we demonstrate that much higher performance can be achieved with architecture space exploration using any optimization algorithms with system level performance and area models for hardware primitives as inputs and goal of minimizing latency with given resource constraints. With this method, ~30GOPs performance was achieved for Stratix V FPGA boards.
Hardware acceleration of DL algorithms alone is not always the most ecient way and sucient to achieve desired performance. There is a huge headroom available for performance improvement provided the algorithms are designed keeping in mind the hardware limitations and bottlenecks. This work achieves hardware-software co-optimization for Non-Maximal Suppression (NMS) algorithm. Using the proposed algorithmic changes and hardware architecture
With CMOS scaling coming to an end and increasing memory bandwidth bottlenecks, CMOS based system might not scale enough to accommodate requirements of more complicated and deeper neural networks in future. In this work, we explore RRAM crossbars and arrays as compact, high performing and energy efficient alternative to CMOS accelerators for deep learning training and inference. We propose and implement RRAM periphery read and write circuits and achieved ~3000x performance improvement in online dictionary learning compared to CPU.
This work also examines the realistic RRAM devices and their non-idealities. We do an in-depth study of the effects of RRAM non-idealities on inference accuracy when a pretrained model is mapped to RRAM based accelerators. To mitigate this issue, we propose Random Sparse Adaptation (RSA), a novel scheme aimed at tuning the model to take care of the faults of the RRAM array on which it is mapped. Our proposed method can achieve inference accuracy much higher than what traditional Read-Verify-Write (R-V-W) method could achieve. RSA can also recover lost inference accuracy 100x ~ 1000x faster compared to R-V-W. Using 32-bit high precision RSA cells, we achieved ~10% higher accuracy using fautly RRAM arrays compared to what can be achieved by mapping a deep network to an 32 level RRAM array with no variations.
Deep neural networks are composed of multiple layers that are compute/memory intensive. This makes it difficult to execute these algorithms real-time with low power consumption using existing general purpose computers. In this work, we propose hardware accelerators for a traditional AI algorithm based on random forest trees and two representative deep convolutional neural networks (AlexNet and VGG). With the proposed acceleration techniques, ~ 30x performance improvement was achieved compared to CPU for random forest trees. For deep CNNS, we demonstrate that much higher performance can be achieved with architecture space exploration using any optimization algorithms with system level performance and area models for hardware primitives as inputs and goal of minimizing latency with given resource constraints. With this method, ~30GOPs performance was achieved for Stratix V FPGA boards.
Hardware acceleration of DL algorithms alone is not always the most ecient way and sucient to achieve desired performance. There is a huge headroom available for performance improvement provided the algorithms are designed keeping in mind the hardware limitations and bottlenecks. This work achieves hardware-software co-optimization for Non-Maximal Suppression (NMS) algorithm. Using the proposed algorithmic changes and hardware architecture
With CMOS scaling coming to an end and increasing memory bandwidth bottlenecks, CMOS based system might not scale enough to accommodate requirements of more complicated and deeper neural networks in future. In this work, we explore RRAM crossbars and arrays as compact, high performing and energy efficient alternative to CMOS accelerators for deep learning training and inference. We propose and implement RRAM periphery read and write circuits and achieved ~3000x performance improvement in online dictionary learning compared to CPU.
This work also examines the realistic RRAM devices and their non-idealities. We do an in-depth study of the effects of RRAM non-idealities on inference accuracy when a pretrained model is mapped to RRAM based accelerators. To mitigate this issue, we propose Random Sparse Adaptation (RSA), a novel scheme aimed at tuning the model to take care of the faults of the RRAM array on which it is mapped. Our proposed method can achieve inference accuracy much higher than what traditional Read-Verify-Write (R-V-W) method could achieve. RSA can also recover lost inference accuracy 100x ~ 1000x faster compared to R-V-W. Using 32-bit high precision RSA cells, we achieved ~10% higher accuracy using fautly RRAM arrays compared to what can be achieved by mapping a deep network to an 32 level RRAM array with no variations.
ContributorsMohanty, Abinash (Author) / Cao, Yu (Thesis advisor) / Seo, Jae-Sun (Committee member) / Vrudhula, Sarma (Committee member) / Chakrabarti, Chaitali (Committee member) / Arizona State University (Publisher)
Created2018
Description
Speech recognition and keyword detection are becoming increasingly popular applications for mobile systems. While deep neural network (DNN) implementation of these systems have very good performance,
they have large memory and compute resource requirements, making their implementation on a mobile device quite challenging. In this thesis, techniques to reduce the memory and computation cost
of keyword detection and speech recognition networks (or DNNs) are presented.
The first technique is based on representing all weights and biases by a small number of bits and mapping all nodal computations into fixed-point ones with minimal degradation in the
accuracy. Experiments conducted on the Resource Management (RM) database show that for the keyword detection neural network, representing the weights by 5 bits results in a 6 fold reduction in memory compared to a floating point implementation with very little loss in performance. Similarly, for the speech recognition neural network, representing the weights by 6 bits results in a 5 fold reduction in memory while maintaining an error rate similar to a floating point implementation. Additional reduction in memory is achieved by a technique called weight pruning,
where the weights are classified as sensitive and insensitive and the sensitive weights are represented with higher precision. A combination of these two techniques helps reduce the memory
footprint by 81 - 84% for speech recognition and keyword detection networks respectively.
Further reduction in memory size is achieved by judiciously dropping connections for large blocks of weights. The corresponding technique, termed coarse-grain sparsification, introduces
hardware-aware sparsity during DNN training, which leads to efficient weight memory compression and significant reduction in the number of computations during classification without
loss of accuracy. Keyword detection and speech recognition DNNs trained with 75% of the weights dropped and classified with 5-6 bit weight precision effectively reduced the weight memory
requirement by ~95% compared to a fully-connected network with double precision, while showing similar performance in keyword detection accuracy and word error rate.
they have large memory and compute resource requirements, making their implementation on a mobile device quite challenging. In this thesis, techniques to reduce the memory and computation cost
of keyword detection and speech recognition networks (or DNNs) are presented.
The first technique is based on representing all weights and biases by a small number of bits and mapping all nodal computations into fixed-point ones with minimal degradation in the
accuracy. Experiments conducted on the Resource Management (RM) database show that for the keyword detection neural network, representing the weights by 5 bits results in a 6 fold reduction in memory compared to a floating point implementation with very little loss in performance. Similarly, for the speech recognition neural network, representing the weights by 6 bits results in a 5 fold reduction in memory while maintaining an error rate similar to a floating point implementation. Additional reduction in memory is achieved by a technique called weight pruning,
where the weights are classified as sensitive and insensitive and the sensitive weights are represented with higher precision. A combination of these two techniques helps reduce the memory
footprint by 81 - 84% for speech recognition and keyword detection networks respectively.
Further reduction in memory size is achieved by judiciously dropping connections for large blocks of weights. The corresponding technique, termed coarse-grain sparsification, introduces
hardware-aware sparsity during DNN training, which leads to efficient weight memory compression and significant reduction in the number of computations during classification without
loss of accuracy. Keyword detection and speech recognition DNNs trained with 75% of the weights dropped and classified with 5-6 bit weight precision effectively reduced the weight memory
requirement by ~95% compared to a fully-connected network with double precision, while showing similar performance in keyword detection accuracy and word error rate.
ContributorsArunachalam, Sairam (Author) / Chakrabarti, Chaitali (Thesis advisor) / Seo, Jae-Sun (Thesis advisor) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2016
Description
Machine learning technology has made a lot of incredible achievements in recent years. It has rivalled or exceeded human performance in many intellectual tasks including image recognition, face detection and the Go game. Many machine learning algorithms require huge amount of computation such as in multiplication of large matrices. As silicon technology has scaled to sub-14nm regime, simply scaling down the device cannot provide enough speed-up any more. New device technologies and system architectures are needed to improve the computing capacity. Designing specific hardware for machine learning is highly in demand. Efforts need to be made on a joint design and optimization of both hardware and algorithm.
For machine learning acceleration, traditional SRAM and DRAM based system suffer from low capacity, high latency, and high standby power. Instead, emerging memories, such as Phase Change Random Access Memory (PRAM), Spin-Transfer Torque Magnetic Random Access Memory (STT-MRAM), and Resistive Random Access Memory (RRAM), are promising candidates providing low standby power, high data density, fast access and excellent scalability. This dissertation proposes a hierarchical memory modeling framework and models PRAM and STT-MRAM in four different levels of abstraction. With the proposed models, various simulations are conducted to investigate the performance, optimization, variability, reliability, and scalability.
Emerging memory devices such as RRAM can work as a 2-D crosspoint array to speed up the multiplication and accumulation in machine learning algorithms. This dissertation proposes a new parallel programming scheme to achieve in-memory learning with RRAM crosspoint array. The programming circuitry is designed and simulated in TSMC 65nm technology showing 900X speedup for the dictionary learning task compared to the CPU performance.
From the algorithm perspective, inspired by the high accuracy and low power of the brain, this dissertation proposes a bio-plausible feedforward inhibition spiking neural network with Spike-Rate-Dependent-Plasticity (SRDP) learning rule. It achieves more than 95% accuracy on the MNIST dataset, which is comparable to the sparse coding algorithm, but requires far fewer number of computations. The role of inhibition in this network is systematically studied and shown to improve the hardware efficiency in learning.
For machine learning acceleration, traditional SRAM and DRAM based system suffer from low capacity, high latency, and high standby power. Instead, emerging memories, such as Phase Change Random Access Memory (PRAM), Spin-Transfer Torque Magnetic Random Access Memory (STT-MRAM), and Resistive Random Access Memory (RRAM), are promising candidates providing low standby power, high data density, fast access and excellent scalability. This dissertation proposes a hierarchical memory modeling framework and models PRAM and STT-MRAM in four different levels of abstraction. With the proposed models, various simulations are conducted to investigate the performance, optimization, variability, reliability, and scalability.
Emerging memory devices such as RRAM can work as a 2-D crosspoint array to speed up the multiplication and accumulation in machine learning algorithms. This dissertation proposes a new parallel programming scheme to achieve in-memory learning with RRAM crosspoint array. The programming circuitry is designed and simulated in TSMC 65nm technology showing 900X speedup for the dictionary learning task compared to the CPU performance.
From the algorithm perspective, inspired by the high accuracy and low power of the brain, this dissertation proposes a bio-plausible feedforward inhibition spiking neural network with Spike-Rate-Dependent-Plasticity (SRDP) learning rule. It achieves more than 95% accuracy on the MNIST dataset, which is comparable to the sparse coding algorithm, but requires far fewer number of computations. The role of inhibition in this network is systematically studied and shown to improve the hardware efficiency in learning.
ContributorsXu, Zihan (Author) / Cao, Yu (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Seo, Jae-Sun (Committee member) / Yu, Shimeng (Committee member) / Arizona State University (Publisher)
Created2017
Description
The advances of Deep Learning (DL) achieved recently have successfully demonstrated its great potential of surpassing or close to human-level performance across multiple domains. Consequently, there exists a rising demand to deploy state-of-the-art DL algorithms, e.g., Deep Neural Networks (DNN), in real-world applications to release labors from repetitive work. On the one hand, the impressive performance achieved by the DNN normally accompanies with the drawbacks of intensive memory and power usage due to enormous model size and high computation workload, which significantly hampers their deployment on the resource-limited cyber-physical systems or edge devices. Thus, the urgent demand for enhancing the inference efficiency of DNN has also great research interests across various communities. On the other hand, scientists and engineers still have insufficient knowledge about the principles of DNN which makes it mostly be treated as a black-box. Under such circumstance, DNN is like "the sword of Damocles" where its security or fault-tolerance capability is an essential concern which cannot be circumvented.
Motivated by the aforementioned concerns, this dissertation comprehensively investigates the emerging efficiency and security issues of DNNs, from both software and hardware design perspectives. From the efficiency perspective, as the foundation technique for efficient inference of target DNN, the model compression via quantization is elaborated. In order to maximize the inference performance boost, the deployment of quantized DNN on the revolutionary Computing-in-Memory based neural accelerator is presented in a cross-layer (device/circuit/system) fashion. From the security perspective, the well known adversarial attack is investigated spanning from its original input attack form (aka. Adversarial example generation) to its parameter attack variant.
Motivated by the aforementioned concerns, this dissertation comprehensively investigates the emerging efficiency and security issues of DNNs, from both software and hardware design perspectives. From the efficiency perspective, as the foundation technique for efficient inference of target DNN, the model compression via quantization is elaborated. In order to maximize the inference performance boost, the deployment of quantized DNN on the revolutionary Computing-in-Memory based neural accelerator is presented in a cross-layer (device/circuit/system) fashion. From the security perspective, the well known adversarial attack is investigated spanning from its original input attack form (aka. Adversarial example generation) to its parameter attack variant.
Contributorshe, zhezhi (Author) / Fan, Deliang (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Cao, Yu (Committee member) / Seo, Jae-Sun (Committee member) / Arizona State University (Publisher)
Created2020
Description
Rapid development of computer vision applications such as image recognition and object detection has been enabled by the emerging deep learning technologies. To improve the accuracy further, deeper and wider neural networks with diverse architecture are proposed for better feature extraction. Though the performance boost is impressive, only marginal improvement can be achieved with significantly increased computational overhead. One solution is to compress the exploding-sized model by dropping less important weights or channels. This is an effective solution that has been well explored. However, by utilizing the rich relation information of the data, one can also improve the accuracy with reasonable overhead. This work makes progress toward efficient and accurate visual tasks including detection, prediction and understanding by using relations.
For object detection, a novel approach, Graph Assisted Reasoning (GAR), is proposed to utilize a heterogeneous graph to model object-object relations and object-scene relations. GAR fuses the features from neighboring object nodes as well as scene nodes. In this way, GAR produces better recognition than that produced from individual object nodes. Moreover, compared to previous approaches using Recurrent Neural Network (RNN), GAR's light-weight and low-coupling architecture further facilitate its integration into the object detection module.
For trajectories prediction, a novel approach, namely Diverse Attention RNN (DAT-RNN), is proposed to handle the diversity of trajectories and modeling of neighboring relations. DAT-RNN integrates both temporal and spatial relations to improve the prediction under various circumstances.
Last but not least, this work presents a novel relation implication-enhanced (RIE) approach that improves relation detection through relation direction and implication. With the relation implication, the SGG model is exposed to more ground truth information and thus mitigates the overfitting problem of the biased datasets. Moreover, the enhancement with relation implication is compatible with various context encoding schemes.
Comprehensive experiments on benchmarking datasets demonstrate the efficacy of the proposed approaches.
For object detection, a novel approach, Graph Assisted Reasoning (GAR), is proposed to utilize a heterogeneous graph to model object-object relations and object-scene relations. GAR fuses the features from neighboring object nodes as well as scene nodes. In this way, GAR produces better recognition than that produced from individual object nodes. Moreover, compared to previous approaches using Recurrent Neural Network (RNN), GAR's light-weight and low-coupling architecture further facilitate its integration into the object detection module.
For trajectories prediction, a novel approach, namely Diverse Attention RNN (DAT-RNN), is proposed to handle the diversity of trajectories and modeling of neighboring relations. DAT-RNN integrates both temporal and spatial relations to improve the prediction under various circumstances.
Last but not least, this work presents a novel relation implication-enhanced (RIE) approach that improves relation detection through relation direction and implication. With the relation implication, the SGG model is exposed to more ground truth information and thus mitigates the overfitting problem of the biased datasets. Moreover, the enhancement with relation implication is compatible with various context encoding schemes.
Comprehensive experiments on benchmarking datasets demonstrate the efficacy of the proposed approaches.
ContributorsLi, Zheng (Author) / Cao, Yu (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Seo, Jae-Sun (Committee member) / Fan, Deliang (Committee member) / Arizona State University (Publisher)
Created2020
Description
Artificial intelligence is one of the leading technologies that mimics the problem solving and decision making capabilities of the human brain. Machine learning algorithms, especially deep learning algorithms, are leading the way in terms of performance and robustness. They are used for various purposes, mainly for computer vision, speech recognition, and object detection. The algorithms are usually tested inaccuracy, and they utilize full floating-point precision (32 bits). The hardware would require a high amount of power and area to accommodate many parameters with full precision. In this exploratory work, the convolution autoencoder is quantized for the working of an event base camera. The model is designed so that the autoencoder can work on-chip, which would sufficiently decrease the latency in processing. Different quantization methods are used to quantize and binarize the weights and activations of this neural network model to be portable and power efficient. The sparsity term is added to make the model as robust and energy-efficient as possible. The network model was able to recoup the lost accuracy due to binarizing the weights and activation's to quantize the layers of the encoder selectively. This method of recouping the accuracy gives enough flexibility to introduce the network on the chip to get real-time processing from systems like event-based cameras. Lately, computer vision, especially object detection have made strides in their object detection accuracy. The algorithms can sufficiently detect and predict the objects in real-time. However, end-to-end detection of the algorithm is challenging due to the large parameter need and processing requirements. A change in the Non Maximum Suppression algorithm in SSD(Single Shot Detector)-Mobilenet-V1 resulted in less computational complexity without change in the quality of output metric. The Mean Average Precision(mAP) calculated suggests that this method can be implemented in the post-processing of other networks.
ContributorsKuzhively, Ajay Balu (Author) / Cao, Yu (Thesis advisor) / Seo, Jae-Sun (Committee member) / Fan, Delian (Committee member) / Arizona State University (Publisher)
Created2021