Matching Items (15)
Filtering by
- All Subjects: Computer Engineering
- Genre: Masters Thesis
- Creators: Seo, Jae-Sun
- Creators: Sohoni, Sohum
Description
Each programming language has a compiler associated with it which helps to identify logical or syntactical errors in the program. These compiler error messages play important part in the form of formative feedback for the programmer. Thus, the error messages should be constructed carefully, considering the affective and cognitive needs of programmers. This is especially true for systems that are used in educational settings, as the messages are typically seen by students who are novice programmers. If the error messages are hard to understand then they might discourage students from understanding or learning the programming language. The primary goal of this research is to identify methods to make the error messages more effective so that students can understand them better and simultaneously learn from their mistakes. This study is focused on understanding how the error message affects the understanding of the error and the approach students take to solve the error. In this study, three types of error messages were provided to the students. The first type is Default type error message which is an assembler centric error message. The second type is Link type error message which is a descriptive error message along with a link to the appropriate section of the PLP manual. The third type is Example type error message which is again a descriptive error message with an example of the similar type of error along with correction step. All these error types were developed for the PLP assembly language. A think-aloud experiment was designed and conducted on the students. The experiment was later transcribed and coded to understand different approach students take to solve different type of error message. After analyzing the result of the think-aloud experiment it was found that student read the Link type error message completely and they understood and learned from the error message to solve the error. The results also indicated that Link type was more helpful compare to other types of error message. The Link type made error solving process more effective compared to other error types.
ContributorsTanpure, Siddhant Bapusaheb (Author) / Sohoni, Sohum (Thesis advisor) / Gary, Kevin A (Committee member) / Craig, Scotty D. (Committee member) / Arizona State University (Publisher)
Created2018
Description
Graph theory is a critical component of computer science and software engineering, with algorithms concerning graph traversal and comprehension powering much of the largest problems in both industry and research. Engineers and researchers often have an accurate view of their target graph, however they struggle to implement a correct, and efficient, search over that graph.
To facilitate rapid, correct, efficient, and intuitive development of graph based solutions we propose a new programming language construct - the search statement. Given a supra-root node, a procedure which determines the children of a given parent node, and optional definitions of the fail-fast acceptance or rejection of a solution, the search statement can conduct a search over any graph or network. Structurally, this statement is modelled after the common switch statement and is put into a largely imperative/procedural context to allow for immediate and intuitive development by most programmers. The Go programming language has been used as a foundation and proof-of-concept of the search statement. A Go compiler is provided which implements this construct.
To facilitate rapid, correct, efficient, and intuitive development of graph based solutions we propose a new programming language construct - the search statement. Given a supra-root node, a procedure which determines the children of a given parent node, and optional definitions of the fail-fast acceptance or rejection of a solution, the search statement can conduct a search over any graph or network. Structurally, this statement is modelled after the common switch statement and is put into a largely imperative/procedural context to allow for immediate and intuitive development by most programmers. The Go programming language has been used as a foundation and proof-of-concept of the search statement. A Go compiler is provided which implements this construct.
ContributorsHenderson, Christopher (Author) / Bansal, Ajay (Thesis advisor) / Lindquist, Timothy (Committee member) / Acuna, Ruben (Committee member) / Arizona State University (Publisher)
Created2018
Description
Deep neural networks (DNN) have shown tremendous success in various cognitive tasks, such as image classification, speech recognition, etc. However, their usage on resource-constrained edge devices has been limited due to high computation and large memory requirement.
To overcome these challenges, recent works have extensively investigated model compression techniques such as element-wise sparsity, structured sparsity and quantization. While most of these works have applied these compression techniques in isolation, there have been very few studies on application of quantization and structured sparsity together on a DNN model.
This thesis co-optimizes structured sparsity and quantization constraints on DNN models during training. Specifically, it obtains optimal setting of 2-bit weight and 2-bit activation coupled with 4X structured compression by performing combined exploration of quantization and structured compression settings. The optimal DNN model achieves 50X weight memory reduction compared to floating-point uncompressed DNN. This memory saving is significant since applying only structured sparsity constraints achieves 2X memory savings and only quantization constraints achieves 16X memory savings. The algorithm has been validated on both high and low capacity DNNs and on wide-sparse and deep-sparse DNN models. Experiments demonstrated that deep-sparse DNN outperforms shallow-dense DNN with varying level of memory savings depending on DNN precision and sparsity levels. This work further proposed a Pareto-optimal approach to systematically extract optimal DNN models from a huge set of sparse and dense DNN models. The resulting 11 optimal designs were further evaluated by considering overall DNN memory which includes activation memory and weight memory. It was found that there is only a small change in the memory footprint of the optimal designs corresponding to the low sparsity DNNs. However, activation memory cannot be ignored for high sparsity DNNs.
To overcome these challenges, recent works have extensively investigated model compression techniques such as element-wise sparsity, structured sparsity and quantization. While most of these works have applied these compression techniques in isolation, there have been very few studies on application of quantization and structured sparsity together on a DNN model.
This thesis co-optimizes structured sparsity and quantization constraints on DNN models during training. Specifically, it obtains optimal setting of 2-bit weight and 2-bit activation coupled with 4X structured compression by performing combined exploration of quantization and structured compression settings. The optimal DNN model achieves 50X weight memory reduction compared to floating-point uncompressed DNN. This memory saving is significant since applying only structured sparsity constraints achieves 2X memory savings and only quantization constraints achieves 16X memory savings. The algorithm has been validated on both high and low capacity DNNs and on wide-sparse and deep-sparse DNN models. Experiments demonstrated that deep-sparse DNN outperforms shallow-dense DNN with varying level of memory savings depending on DNN precision and sparsity levels. This work further proposed a Pareto-optimal approach to systematically extract optimal DNN models from a huge set of sparse and dense DNN models. The resulting 11 optimal designs were further evaluated by considering overall DNN memory which includes activation memory and weight memory. It was found that there is only a small change in the memory footprint of the optimal designs corresponding to the low sparsity DNNs. However, activation memory cannot be ignored for high sparsity DNNs.
ContributorsSrivastava, Gaurav (Author) / Seo, Jae-Sun (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Berisha, Visar (Committee member) / Arizona State University (Publisher)
Created2018
Description
Hardware implementation of deep neural networks is earning significant importance nowadays. Deep neural networks are mathematical models that use learning algorithms inspired by the brain. Numerous deep learning algorithms such as multi-layer perceptrons (MLP) have demonstrated human-level recognition accuracy in image and speech classification tasks. Multiple layers of processing elements called neurons with several connections between them called synapses are used to build these networks. Hence, it involves operations that exhibit a high level of parallelism making it computationally and memory intensive. Constrained by computing resources and memory, most of the applications require a neural network which utilizes less energy. Energy efficient implementation of these computationally intense algorithms on neuromorphic hardware demands a lot of architectural optimizations. One of these optimizations would be the reduction in the network size using compression and several studies investigated compression by introducing element-wise or row-/column-/block-wise sparsity via pruning and regularization. Additionally, numerous recent works have concentrated on reducing the precision of activations and weights with some reducing to a single bit. However, combining various sparsity structures with binarized or very-low-precision (2-3 bit) neural networks have not been comprehensively explored. Output activations in these deep neural network algorithms are habitually non-binary making it difficult to exploit sparsity. On the other hand, biologically realistic models like spiking neural networks (SNN) closely mimic the operations in biological nervous systems and explore new avenues for brain-like cognitive computing. These networks deal with binary spikes, and they can exploit the input-dependent sparsity or redundancy to dynamically scale the amount of computation in turn leading to energy-efficient hardware implementation. This work discusses configurable spiking neuromorphic architecture that supports multiple hidden layers exploiting hardware reuse. It also presents design techniques for minimum-area/-energy DNN hardware with minimal degradation in accuracy. Area, performance and energy results of these DNN and SNN hardware is reported for the MNIST dataset. The Neuromorphic hardware designed for SNN algorithm in 28nm CMOS demonstrates high classification accuracy (>98% on MNIST) and low energy (51.4 - 773 (nJ) per classification). The optimized DNN hardware designed in 40nm CMOS that combines 8X structured compression and 3-bit weight precision showed 98.4% accuracy at 33 (nJ) per classification.
ContributorsKolala Venkataramanaiah, Shreyas (Author) / Seo, Jae-Sun (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2018
Description
Integrated circuits must be energy efficient. This efficiency affects all aspects of chip design, from the battery life of embedded devices to thermal heating on high performance servers. As technology scaling slows, future generations of transistors will lack the energy efficiency gains as it has had in previous generations. Therefore, other sources of energy efficiency will be much more important. Many computations have the potential to be executed for extreme energy efficiency but are not instigated because the platforms they run on are not optimized for efficient execution. ASICs improve energy efficiency by reducing flexibility and leveraging the properties of a specific computation. However, ASICs are fixed in function and therefore have incredible opportunity cost. FPGAs offer a reconfigurable solution but are 25x less energy efficient than ASIC implementation. Spatially programmable architectures (SPAs) are similar in design and structure to ASICs and FPGAs but are able bridge the ASIC-FPGA energy efficiency gap by trading flexibility for efficiency. However, SPAs are difficult to program because they do not share the same programming model as normal architectures that execute in time. This work addresses compiler challenges for coarse grained, locally interconnected SPA for domain efficiency (SPADE). A novel SPADE topology, called the wave pipeline, is introduced that is designed for the image signal processing domain that is both efficient and simple to compile to. A compiler for the wave pipeline is created that solves for maximum energy and area efficiency using low complexity, greedy methods. The wave pipeline topology and compiler allow for us to investigate and experiment with image signal processing applications to prove the feasibility of SPADE compilers.
ContributorsMackay, Curtis (Author) / Brunhaver, John (Thesis advisor) / Karam, Lina J (Committee member) / Seo, Jae-Sun (Committee member) / Arizona State University (Publisher)
Created2016
Description
The technological advances in the past few decades have made possible creation and consumption of digital visual content at an explosive rate. Consequently, there is a need for efficient quality monitoring systems to ensure minimal degradation of images and videos during various processing operations like compression, transmission, storage etc. Objective Image Quality Assessment (IQA) algorithms have been developed that predict quality scores which match well with human subjective quality assessment. However, a lot of research still remains to be done before IQA algorithms can be deployed in real world systems. Long runtimes for one frame of image is a major hurdle. Graphics Processing Units (GPUs), equipped with massive number of computational cores, provide an opportunity to accelerate IQA algorithms by performing computations in parallel. Indeed, General Purpose Graphics Processing Units (GPGPU) techniques have been applied to a few Full Reference IQA algorithms which fall under the. We present a GPGPU implementation of Blind Image Integrity Notator using DCT Statistics (BLIINDS-II), which falls under the No Reference IQA algorithm paradigm. We have been able to achieve a speedup of over 30x over the previous CPU version of this algorithm. We test our implementation using various distorted images from the CSIQ database and present the performance trends observed. We achieve a very consistent performance of around 9 milliseconds per distorted image, which made possible the execution of over 100 images per second (100 fps).
ContributorsYadav, Aman (Author) / Sohoni, Sohum (Thesis advisor) / Aukes, Daniel (Committee member) / Redkar, Sangram (Committee member) / Arizona State University (Publisher)
Created2016
Description
With the advent of GPGPU, many applications are being accelerated by using CUDA programing paradigm. We are able to achieve around 10x -100x speedups by simply porting the application on to the GPU and running the parallel chunk of code on its multi cored SIMT (Single instruction multiple thread) architecture. But for optimal performance it is necessary to make sure that all the GPU resources are efficiently used, and the latencies in the application are minimized. For this, it is essential to monitor the Hardware usage of the algorithm and thus diagnose the compute and memory bottlenecks in the implementation. In the following thesis, we will be analyzing the mapping of CUDA implementation of BLIINDS-II algorithm on the underlying GPU hardware, and come up with a Kepler architecture specific solution of using shuffle instruction via CUB library to tackle the two major bottlenecks in the algorithm. Experiments were conducted to convey the advantage of using shuffle instru3ction in algorithm over only using shared memory as a buffer to global memory. With the new implementation of BLIINDS-II algorithm using CUB library, a speedup of around 13.7% was achieved.
ContributorsWadekar, Ameya (Author) / Sohoni, Sohum (Thesis advisor) / Aukes, Daniel (Committee member) / Redkar, Sangram (Committee member) / Arizona State University (Publisher)
Created2017
Description
The information era has brought about many technological advancements in the past
few decades, and that has led to an exponential increase in the creation of digital images and
videos. Constantly, all digital images go through some image processing algorithm for
various reasons like compression, transmission, storage, etc. There is data loss during this
process which leaves us with a degraded image. Hence, to ensure minimal degradation of
images, the requirement for quality assessment has become mandatory. Image Quality
Assessment (IQA) has been researched and developed over the last several decades to
predict the quality score in a manner that agrees with human judgments of quality. Modern
image quality assessment (IQA) algorithms are quite effective at prediction accuracy, and
their development has not focused on improving computational performance. The existing
serial implementation requires a relatively large run-time on the order of seconds for a single
frame. Hardware acceleration using Field programmable gate arrays (FPGAs) provides
reconfigurable computing fabric that can be tailored for a broad range of applications.
Usually, programming FPGAs has required expertise in hardware descriptive languages
(HDLs) or high-level synthesis (HLS) tool. OpenCL is an open standard for cross-platform,
parallel programming of heterogeneous systems along with Altera OpenCL SDK, enabling
developers to use FPGA's potential without extensive hardware knowledge. Hence, this
thesis focuses on accelerating the computationally intensive part of the most apparent
distortion (MAD) algorithm on FPGA using OpenCL. The results are compared with CPU
implementation to evaluate performance and efficiency gains.
few decades, and that has led to an exponential increase in the creation of digital images and
videos. Constantly, all digital images go through some image processing algorithm for
various reasons like compression, transmission, storage, etc. There is data loss during this
process which leaves us with a degraded image. Hence, to ensure minimal degradation of
images, the requirement for quality assessment has become mandatory. Image Quality
Assessment (IQA) has been researched and developed over the last several decades to
predict the quality score in a manner that agrees with human judgments of quality. Modern
image quality assessment (IQA) algorithms are quite effective at prediction accuracy, and
their development has not focused on improving computational performance. The existing
serial implementation requires a relatively large run-time on the order of seconds for a single
frame. Hardware acceleration using Field programmable gate arrays (FPGAs) provides
reconfigurable computing fabric that can be tailored for a broad range of applications.
Usually, programming FPGAs has required expertise in hardware descriptive languages
(HDLs) or high-level synthesis (HLS) tool. OpenCL is an open standard for cross-platform,
parallel programming of heterogeneous systems along with Altera OpenCL SDK, enabling
developers to use FPGA's potential without extensive hardware knowledge. Hence, this
thesis focuses on accelerating the computationally intensive part of the most apparent
distortion (MAD) algorithm on FPGA using OpenCL. The results are compared with CPU
implementation to evaluate performance and efficiency gains.
ContributorsGunavelu Mohan, Aswin (Author) / Sohoni, Sohum (Thesis advisor) / Ren, Fengbo (Thesis advisor) / Seo, Jae-Sun (Committee member) / Arizona State University (Publisher)
Created2017
Description
Image processing has changed the way we store, view and share images. One important component of sharing images over the networks is image compression. Lossy image compression techniques compromise the quality of images to reduce their size. To ensure that the distortion of images due to image compression is not highly detectable by humans, the perceived quality of an image needs to be maintained over a certain threshold. Determining this threshold is best done using human subjects, but that is impractical in real-world scenarios. As a solution to this issue, image quality assessment (IQA) algorithms are used to automatically compute a fidelity score of an image.
However, poor performance of IQA algorithms has been observed due to complex statistical computations involved. General Purpose Graphics Processing Unit (GPGPU) programming is one of the solutions proposed to optimize the performance of these algorithms.
This thesis presents a Compute Unified Device Architecture (CUDA) based optimized implementation of full reference IQA algorithm, Visual Signal to Noise Ratio (VSNR) that uses M-level 2D Discrete Wavelet Transform (DWT) with 9/7 biorthogonal filters among other statistical computations. The presented implementation is tested upon four different image quality databases containing images with multiple distortions and sizes ranging from 512 x 512 to 1600 x 1280. The CUDA implementation of VSNR shows a speedup of over 32x for 1600 x 1280 images. It is observed that the speedup scales with the increase in size of images. The results showed that the implementation is fast enough to use VSNR on high definition videos with a frame rate of 60 fps. This work presents the optimizations made due to the use of GPU’s constant memory and reuse of allocated memory on the GPU. Also, it shows the performance improvement using profiler driven GPGPU development in CUDA. The presented implementation can be deployed in production combined with existing applications.
However, poor performance of IQA algorithms has been observed due to complex statistical computations involved. General Purpose Graphics Processing Unit (GPGPU) programming is one of the solutions proposed to optimize the performance of these algorithms.
This thesis presents a Compute Unified Device Architecture (CUDA) based optimized implementation of full reference IQA algorithm, Visual Signal to Noise Ratio (VSNR) that uses M-level 2D Discrete Wavelet Transform (DWT) with 9/7 biorthogonal filters among other statistical computations. The presented implementation is tested upon four different image quality databases containing images with multiple distortions and sizes ranging from 512 x 512 to 1600 x 1280. The CUDA implementation of VSNR shows a speedup of over 32x for 1600 x 1280 images. It is observed that the speedup scales with the increase in size of images. The results showed that the implementation is fast enough to use VSNR on high definition videos with a frame rate of 60 fps. This work presents the optimizations made due to the use of GPU’s constant memory and reuse of allocated memory on the GPU. Also, it shows the performance improvement using profiler driven GPGPU development in CUDA. The presented implementation can be deployed in production combined with existing applications.
ContributorsGupta, Ayush (Author) / Sohoni, Sohum (Thesis advisor) / Amresh, Ashish (Committee member) / Bansal, Ajay (Committee member) / Arizona State University (Publisher)
Created2017
Description
Achieving human level intelligence is a long-term goal for many Artificial Intelligence (AI) researchers. Recent developments in combining deep learning and reinforcement learning helped us to move a step forward in achieving this goal. Reinforcement learning using a delayed reward mechanism is an approach to machine intelligence which studies decision making with control and how a decision making agent can learn to act optimally in an environment-unaware conditions.
Q-learning is one of the model-free reinforcement directed learning strategies which uses temporal differences to estimate the performances of state-action pairs called Q values. A simple implementation of Q-learning algorithm can be done using a Q table memory to store and update the Q values. However, with an increase in state space data due to a complex environment, and with an increase in possible number of actions an agent can perform, Q table reaches its space limit and would be difficult to scale well. Q-learning with neural networks eliminates the use of Q table by approximating the Q function using neural networks.
Autonomous agents need to develop cognitive properties and become self-adaptive to be deployable in any environment. Reinforcement learning with Q-learning have been very efficient in solving such problems. However, embedded systems like space rovers and autonomous robots rarely implement such techniques due to the constraints faced like processing power, chip area, convergence rate and cost of the chip. These problems present a need for a portable, low power, area efficient hardware accelerator to accelerate the process of such learning.
This problem is targeted by implementing a hardware schematic architecture for Q-learning using Artificial Neural networks. This architecture exploits the massive parallelism provided by neural network with a dedicated fine grain parallelism provided by a Field Programmable Gate Array (FPGA) thereby processing the Q values at a high throughput. Mars exploration rovers currently use Xilinx-Space-grade FPGA devices for image processing, pyrotechnic operation control and obstacle avoidance. The hardware resource consumption for the architecture has been synthesized considering Xilinx Virtex7 FPGA as the target device.
Q-learning is one of the model-free reinforcement directed learning strategies which uses temporal differences to estimate the performances of state-action pairs called Q values. A simple implementation of Q-learning algorithm can be done using a Q table memory to store and update the Q values. However, with an increase in state space data due to a complex environment, and with an increase in possible number of actions an agent can perform, Q table reaches its space limit and would be difficult to scale well. Q-learning with neural networks eliminates the use of Q table by approximating the Q function using neural networks.
Autonomous agents need to develop cognitive properties and become self-adaptive to be deployable in any environment. Reinforcement learning with Q-learning have been very efficient in solving such problems. However, embedded systems like space rovers and autonomous robots rarely implement such techniques due to the constraints faced like processing power, chip area, convergence rate and cost of the chip. These problems present a need for a portable, low power, area efficient hardware accelerator to accelerate the process of such learning.
This problem is targeted by implementing a hardware schematic architecture for Q-learning using Artificial Neural networks. This architecture exploits the massive parallelism provided by neural network with a dedicated fine grain parallelism provided by a Field Programmable Gate Array (FPGA) thereby processing the Q values at a high throughput. Mars exploration rovers currently use Xilinx-Space-grade FPGA devices for image processing, pyrotechnic operation control and obstacle avoidance. The hardware resource consumption for the architecture has been synthesized considering Xilinx Virtex7 FPGA as the target device.
ContributorsGankidi, Pranay Reddy (Author) / Thangavelautham, Jekanthan (Thesis advisor) / Ren, Fengbo (Committee member) / Seo, Jae-Sun (Committee member) / Arizona State University (Publisher)
Created2016