Matching Items (177)
Filtering by
- Genre: Academic theses
- Creators: Vittal, Vijay
- Creators: Ryan, Russell
- Status: Published
Description
This thesis presents efficient implementations of several linear algebra kernels, machine learning kernels and a neural network based recommender systems engine onto a massively parallel reconfigurable architecture, Transformer. The linear algebra kernels include Triangular Matrix Solver (TRSM), LU Decomposition (LUD), QR Decomposition (QRD), and Matrix Inversion. The machine learning kernels include an LSTM (Long Short Term Memory) cell, and a GRU (gated Recurrent Unit) cell used in recurrent neural networks. The neural network based recommender systems engine consists of multiple kernels including fully connected layers, embedding layer, 1-D batchnorm, Adam optimizer, etc.
Transformer is a massively parallel reconfigurable multicore architecture designed at the University of Michigan. The Transformer configuration considered here is 4 tiles and 16 General Processing Elements (GPEs) per tile. It supports a two level cache hierarchy where the L1 and L2 caches can operate in shared (S) or private (P) modes. The architecture was modeled using Gem5 and cycle accurate simulations were done to evaluate the performance in terms of execution times, giga-operations per second per Watt (GOPS/W), and giga-floating-point-operations per second per Watt (GFLOPS/W).
This thesis shows that for linear algebra kernels, each kernel achieves high performance for a certain cache mode and that this cache mode can change when the matrix size changes. For instance, for smaller matrix sizes, L1P, L2P cache mode is best for TRSM, while L1S, L2S is the best cache mode for LUD, and L1P, L2S is the best for QRD. For each kernel, the optimal cache mode changes when the matrix size is increased. For instance, for TRSM, the L1P, L2P cache mode is best for smaller matrix sizes ($N=64, 128, 256, 512$) and it changes to L1S, L2P for larger matrix sizes ($N=1024$). For machine learning kernels, L1P, L2P is the best cache mode for all network parameter sizes.
Gem5 simulations show that the peak performance for TRSM, LUD, QRD and Matrix Inverse in the 14nm node is 97.5, 59.4, 133.0 and 83.05 GFLOPS/W, respectively. For LSTM and GRU, the peak performance is 44.06 and 69.3 GFLOPS/W.
The neural network based recommender system was implemented in L1S, L2S cache mode. It includes a forward pass and a backward pass and is significantly more complex in terms of both computational complexity and data movement. The most computationally intensive block is the fully connected layer followed by Adam optimizer. The overall performance of the recommender systems engine is 54.55 GFLOPS/W and 169.12 GOPS/W.
Transformer is a massively parallel reconfigurable multicore architecture designed at the University of Michigan. The Transformer configuration considered here is 4 tiles and 16 General Processing Elements (GPEs) per tile. It supports a two level cache hierarchy where the L1 and L2 caches can operate in shared (S) or private (P) modes. The architecture was modeled using Gem5 and cycle accurate simulations were done to evaluate the performance in terms of execution times, giga-operations per second per Watt (GOPS/W), and giga-floating-point-operations per second per Watt (GFLOPS/W).
This thesis shows that for linear algebra kernels, each kernel achieves high performance for a certain cache mode and that this cache mode can change when the matrix size changes. For instance, for smaller matrix sizes, L1P, L2P cache mode is best for TRSM, while L1S, L2S is the best cache mode for LUD, and L1P, L2S is the best for QRD. For each kernel, the optimal cache mode changes when the matrix size is increased. For instance, for TRSM, the L1P, L2P cache mode is best for smaller matrix sizes ($N=64, 128, 256, 512$) and it changes to L1S, L2P for larger matrix sizes ($N=1024$). For machine learning kernels, L1P, L2P is the best cache mode for all network parameter sizes.
Gem5 simulations show that the peak performance for TRSM, LUD, QRD and Matrix Inverse in the 14nm node is 97.5, 59.4, 133.0 and 83.05 GFLOPS/W, respectively. For LSTM and GRU, the peak performance is 44.06 and 69.3 GFLOPS/W.
The neural network based recommender system was implemented in L1S, L2S cache mode. It includes a forward pass and a backward pass and is significantly more complex in terms of both computational complexity and data movement. The most computationally intensive block is the fully connected layer followed by Adam optimizer. The overall performance of the recommender systems engine is 54.55 GFLOPS/W and 169.12 GOPS/W.
ContributorsSoorishetty, Anuraag (Author) / Chakrabarti, Chaitali (Thesis advisor) / Kim, Hun Seok (Committee member) / LiKamWa, Robert (Committee member) / Arizona State University (Publisher)
Created2019
Description
Collaborative piano education tends to discuss techniques of collaboration as primarily a musical skill. However, common understanding within the field regarding a collaborative pianist’s ability to work with others offers another aspect to this assumption. It goes without saying that pianists’ interpersonal skills largely affect with whom they will work, and how efficaciously pianists and their partners will work together. Correspondingly, how pianists work with others can directly affect the success or failure of the musical collaboration.
The first intention of this paper is to explain why interpersonal skills are integral to the creation of quality musical outcomes and so-called musical togetherness; it specifies interpersonal aspects innate and unique to a pianist’s experience. Next, this paper defines two crucial components of collaboration – empathy and active listening – and discusses how pianists can build these skills into their personal practice and rehearsal. It continues with an examination of the interpersonal implications of studio arrangement, body language, and verbal language from a pianist’s perspective. This paper concludes with ideas for how to test for these skills during the collaborative piano audition process, a class syllabus showing how these skills can be incorporated into the collaborative piano curriculum, and suggestions for further research about interpersonal aspects of collaboration.
The first intention of this paper is to explain why interpersonal skills are integral to the creation of quality musical outcomes and so-called musical togetherness; it specifies interpersonal aspects innate and unique to a pianist’s experience. Next, this paper defines two crucial components of collaboration – empathy and active listening – and discusses how pianists can build these skills into their personal practice and rehearsal. It continues with an examination of the interpersonal implications of studio arrangement, body language, and verbal language from a pianist’s perspective. This paper concludes with ideas for how to test for these skills during the collaborative piano audition process, a class syllabus showing how these skills can be incorporated into the collaborative piano curriculum, and suggestions for further research about interpersonal aspects of collaboration.
ContributorsCota, Mary Strobel (Author) / Campbell, Andrew (Thesis advisor) / Ryan, Russell (Committee member) / Jiang, Danwen (Committee member) / Feisst, Sabine (Committee member) / Arizona State University (Publisher)
Created2019
Description
The availability of data for monitoring and controlling the electrical grid has increased exponentially over the years in both resolution and quantity leaving a large data footprint. This dissertation is motivated by the need for equivalent representations of grid data in lower-dimensional feature spaces so that machine learning algorithms can be employed for a variety of purposes. To achieve that, without sacrificing the interpretation of the results, the dissertation leverages the physics behind power systems, well-known laws that underlie this man-made infrastructure, and the nature of the underlying stochastic phenomena that define the system operating conditions as the backbone for modeling data from the grid.
The first part of the dissertation introduces a new framework of graph signal processing (GSP) for the power grid, Grid-GSP, and applies it to voltage phasor measurements that characterize the overall system state of the power grid. Concepts from GSP are used in conjunction with known power system models in order to highlight the low-dimensional structure in data and present generative models for voltage phasors measurements. Applications such as identification of graphical communities, network inference, interpolation of missing data, detection of false data injection attacks and data compression are explored wherein Grid-GSP based generative models are used.
The second part of the dissertation develops a model for a joint statistical description of solar photo-voltaic (PV) power and the outdoor temperature which can lead to better management of power generation resources so that electricity demand such as air conditioning and supply from solar power are always matched in the face of stochasticity. The low-rank structure inherent in solar PV power data is used for forecasting and to detect partial-shading type of faults in solar panels.
The first part of the dissertation introduces a new framework of graph signal processing (GSP) for the power grid, Grid-GSP, and applies it to voltage phasor measurements that characterize the overall system state of the power grid. Concepts from GSP are used in conjunction with known power system models in order to highlight the low-dimensional structure in data and present generative models for voltage phasors measurements. Applications such as identification of graphical communities, network inference, interpolation of missing data, detection of false data injection attacks and data compression are explored wherein Grid-GSP based generative models are used.
The second part of the dissertation develops a model for a joint statistical description of solar photo-voltaic (PV) power and the outdoor temperature which can lead to better management of power generation resources so that electricity demand such as air conditioning and supply from solar power are always matched in the face of stochasticity. The low-rank structure inherent in solar PV power data is used for forecasting and to detect partial-shading type of faults in solar panels.
ContributorsRamakrishna, Raksha (Author) / Scaglione, Anna (Thesis advisor) / Cochran, Douglas (Committee member) / Spanias, Andreas (Committee member) / Vittal, Vijay (Committee member) / Zhang, Junshan (Committee member) / Arizona State University (Publisher)
Created2020
Description
This project serves as a performance guide for Chen Yi’s work From Old Peking Folklore for violin and piano. The primary source material for the document is derived from six hours of interviews and musical coaching that the writer undertook in March 2020 with Chen Yi at her residence in Missouri. The work is heavily influenced by Chinese Opera, and a brief examination of the history of Chinese Opera is included to provide context to the performer. Elements of performance practice on traditional Chinese instruments and their influence on the work are also explored, with detailed explanations given for the realization of numerous indications in the score from the composer. Finally, a link to a lecture recital and performance of the work is provided by the writer.
ContributorsDu, Pan (Author) / McLin, Katherine E (Thesis advisor) / Norton, Kay (Committee member) / Ryan, Russell (Committee member) / Arizona State University (Publisher)
Created2020
Description
ABSTRACT
As a composer, Chou Wen-Chung (1923-2019) was a learner and inheritor of Chinese traditional music culture and was committed to carrying it forward. As a native of China who had his primary musical training in the West, Chou Wen-Chung was one of the first Chinese composers to make his mark on Western music. He successfully combined Western elements and Chinese tradition in his music. Chou Wen-Chung was one of the few prominent East Asian composers known in the Western musical world, and his music therefore has had a strong influence on other Chinese composers.
In order to understand more clearly his music, I analyzed his chamber work: Yü Ko. This piece was composed in 1965 for 9 instruments: Violin, Alto Flute, English Horn, Bass Clarinet, 2 Trombones, 2 Percussion and Piano. Inspired by the ancient Chinese musical instrument the Qin (also called guqin, or “ancient qin”), which is a plucked seven-string instrument, Chou Wen-Chung composed Yü Ko. Literally meaning “fisherman’s song,” this work was composed originally for the Qin, based on a melody composed by Mao Min-Zhong who was a very noted scholar and Qin player of the late Southern Song dynasty (C.E.1127-1276).
This paper provides Chou Wen-Chung’s biography, compositional styles and developments. It lists and explains the most common Chinese traditional cultural elements which he used in his compositions. In particular, it introduces the Qin in detail from the external structure, performance techniques, sound characteristics, the tablature notation, and compositional methods.
This document also includes a detailed analysis of Yü Ko in terms of the orchestration, pitch, tonal material, structure and tempo, dynamic and musical materials, and explains Chou Wen-Chung’s imitation of the Qin as well as the influence of Western music shown in this piece.
As a composer, Chou Wen-Chung (1923-2019) was a learner and inheritor of Chinese traditional music culture and was committed to carrying it forward. As a native of China who had his primary musical training in the West, Chou Wen-Chung was one of the first Chinese composers to make his mark on Western music. He successfully combined Western elements and Chinese tradition in his music. Chou Wen-Chung was one of the few prominent East Asian composers known in the Western musical world, and his music therefore has had a strong influence on other Chinese composers.
In order to understand more clearly his music, I analyzed his chamber work: Yü Ko. This piece was composed in 1965 for 9 instruments: Violin, Alto Flute, English Horn, Bass Clarinet, 2 Trombones, 2 Percussion and Piano. Inspired by the ancient Chinese musical instrument the Qin (also called guqin, or “ancient qin”), which is a plucked seven-string instrument, Chou Wen-Chung composed Yü Ko. Literally meaning “fisherman’s song,” this work was composed originally for the Qin, based on a melody composed by Mao Min-Zhong who was a very noted scholar and Qin player of the late Southern Song dynasty (C.E.1127-1276).
This paper provides Chou Wen-Chung’s biography, compositional styles and developments. It lists and explains the most common Chinese traditional cultural elements which he used in his compositions. In particular, it introduces the Qin in detail from the external structure, performance techniques, sound characteristics, the tablature notation, and compositional methods.
This document also includes a detailed analysis of Yü Ko in terms of the orchestration, pitch, tonal material, structure and tempo, dynamic and musical materials, and explains Chou Wen-Chung’s imitation of the Qin as well as the influence of Western music shown in this piece.
ContributorsSong, Yiqian (Author) / Ryan, Russell (Thesis advisor) / Campbell, Andrew (Committee member) / Solís, Ted (Committee member) / Arizona State University (Publisher)
Created2020
Description
The purpose of this document is to create a template for a master’s degree in Collaborative Piano using data collected from an online survey and from publicly available information on institutional websites. The history and development of the graduate collaborative piano degree in the United States is examined to provide the background to this research. In addition to the degree template, other aspects useful for the creation of such a degree are discussed, including proposed required and optional courses, financial considerations, community outreach opportunities, and balancing off-campus professional engagements with on-campus academic duties. A list of all institutions currently offering collaborative piano degrees at the graduate level is included in the appendix.
The degree template conforms to the requirements of the National Association of Schools of Music (NASM) in order to allow the greatest number of institutions the ability to embrace the curriculum. Designed to be flexible within the overall requirements of the degree, the proposed curriculum favors a balanced approach between instrumental and vocal collaboration, with a combination of traditional courses, project-based courses, and customizable elective courses designed to develop important competencies in collaborative piano. Both solo and collaborative applied lessons would be required, with three coached collaborative recitals and one uncoached collaborative recital required to fulfill the degree requirements. The project-oriented Collaborative Piano Seminar course has the flexibility to allow team teaching or community partnerships and requires an off-campus class performance once per academic year.
The goal of this template is to provide a pedagogically solid foundation for a master’s degree in collaborative piano, with the flexibility to add a variety of elective courses best suited to the needs and talents of the students, faculty, and institution. The synthesis of classical and popular styles within the curriculum is designed to give the collaborative pianist diverse musical competencies in order to succeed and thrive as a professional musician in the 21st century, whether the student continues with self-education after the master’s degree, pursues further study at the doctoral level, or enters the professional world.
The degree template conforms to the requirements of the National Association of Schools of Music (NASM) in order to allow the greatest number of institutions the ability to embrace the curriculum. Designed to be flexible within the overall requirements of the degree, the proposed curriculum favors a balanced approach between instrumental and vocal collaboration, with a combination of traditional courses, project-based courses, and customizable elective courses designed to develop important competencies in collaborative piano. Both solo and collaborative applied lessons would be required, with three coached collaborative recitals and one uncoached collaborative recital required to fulfill the degree requirements. The project-oriented Collaborative Piano Seminar course has the flexibility to allow team teaching or community partnerships and requires an off-campus class performance once per academic year.
The goal of this template is to provide a pedagogically solid foundation for a master’s degree in collaborative piano, with the flexibility to add a variety of elective courses best suited to the needs and talents of the students, faculty, and institution. The synthesis of classical and popular styles within the curriculum is designed to give the collaborative pianist diverse musical competencies in order to succeed and thrive as a professional musician in the 21st century, whether the student continues with self-education after the master’s degree, pursues further study at the doctoral level, or enters the professional world.
ContributorsFincher, Aimee Elisabeth (Author) / Campbell, Andrew (Thesis advisor) / Rockmaker, Jody (Committee member) / Ryan, Russell (Committee member) / Schildkret, David (Committee member) / Yeo, Douglas (Committee member) / Arizona State University (Publisher)
Created2020
Description
The work of collaborative pianists can vary widely, requiring a large spectrum of musical and foreign language skills. In addition, many non-musical skills are required of collaborative pianists in order to adapt to various types of work, the roles they assume, and the needs of the people they encounter professionally. Collaborative pianists usually develop good habits for survival on the job, but rarely receive preliminary training in capacities such as facilitation, maintaining objectivity in collaboration, asking good questions, and giving feedback effectively. The emerging field of teaching artistry offers a wealth of information for the development of these non-musical skills in collaborative pianists. The skills necessary for teaching artistry and collaborative piano frequently overlap, which is instructive for collaborative pianists as they prepare for their various musical and leadership roles. This paper explores shared practices between these disciplines, how they can enhance the activities of a collaborative pianist, and also help them develop skills as arts advocates. Advocating techniques for new music and audience engagement are addressed, as well as programming, content development and building teams around projects. The idea of the collaborative pianist becoming a teaching artist is also explored, as the diverse activities and experiences of a collaborative pianist can serve as valuable resources. All of these approaches to non-musical skills focus on building strong processes, leading to creative activities that are process-driven rather than product-driven. This study seeks to enrich activities of collaborative pianists through the application of teaching artist capacities and pave pathways for new, more effective professional collaborations.
ContributorsWitt, Juliana (Author) / Campbell, Andrew (Thesis advisor) / Norton, Kay (Committee member) / Ryan, Russell (Committee member) / Swoboda, Deanna (Committee member) / Arizona State University (Publisher)
Created2020
Description
The electric power system (EPS) is an extremely complex system that has operational interdependencies with the water delivery and treatment system (WDTS). The term water-energy nexus is commonly used to describe the critical interdependencies that naturally exist between the EPS and water distribution systems (WDS). Presented in this work is a framework for simulating interactions between these two critical infrastructure systems in short-term and long-term time-scales. This includes appropriate mathematical models for system modeling and for optimizing control of power system operation with consideration of conditions in the WDS. Also presented is a complete methodology for quantifying the resilience of the two interdependent systems.
The key interdependencies between the two systems are the requirements of water for the cooling cycle of traditional thermal power plants as well as electricity for pumping and/or treatment in the WDS. While previous work has considered the dependency of thermoelectric generation on cooling water requirements at a high-level, this work considers the impact from limitations of cooling water into network simulations in both a short-term operational framework as well as in the long-term planning domain.
The work completed to set-up simulations in operational length time-scales was the development of a simulator that adequately models both systems. This simulation engine also facilitates the implementation of control schemes in both systems that take advantage of the knowledge of operating conditions in the other system. Initial steps for including the influence of anticipated water availability and water rights attainability within the combined generation and transmission expansion planning problem is also presented. Lastly, the framework for determining the infrastructural-operational resilience (IOR) of the interdependent systems is formulated.
Adequately modeling and studying the two systems and their interactions is becoming critically important. This importance is illustrated by the possibility of unforeseen natural or man-made events or by the likelihood of load increase in the systems, either of which has the risk of putting extreme stress on the systems beyond that experienced in normal operating conditions. Therefore, this work addresses these concerns with novel modeling and control/policy strategies designed to mitigate the severity of extreme conditions in either system.
The key interdependencies between the two systems are the requirements of water for the cooling cycle of traditional thermal power plants as well as electricity for pumping and/or treatment in the WDS. While previous work has considered the dependency of thermoelectric generation on cooling water requirements at a high-level, this work considers the impact from limitations of cooling water into network simulations in both a short-term operational framework as well as in the long-term planning domain.
The work completed to set-up simulations in operational length time-scales was the development of a simulator that adequately models both systems. This simulation engine also facilitates the implementation of control schemes in both systems that take advantage of the knowledge of operating conditions in the other system. Initial steps for including the influence of anticipated water availability and water rights attainability within the combined generation and transmission expansion planning problem is also presented. Lastly, the framework for determining the infrastructural-operational resilience (IOR) of the interdependent systems is formulated.
Adequately modeling and studying the two systems and their interactions is becoming critically important. This importance is illustrated by the possibility of unforeseen natural or man-made events or by the likelihood of load increase in the systems, either of which has the risk of putting extreme stress on the systems beyond that experienced in normal operating conditions. Therefore, this work addresses these concerns with novel modeling and control/policy strategies designed to mitigate the severity of extreme conditions in either system.
ContributorsZuloaga, Scott (Author) / Vittal, Vijay (Thesis advisor) / Zhang, Junshan (Committee member) / Mays, Larry (Committee member) / Wu, Meng (Committee member) / Arizona State University (Publisher)
Created2020
Description
Power systems are undergoing a significant transformation as a result of the retirements of conventional coal-fired generation units and the increasing integration of converter interfaced renewable resources. The instantaneous renewable generation penetration as a percentage of the load served in megawatt (MW), in some areas of the United States (U.S.) sometimes approaches over 50 percent. These changes have introduced new challenges for reliability studies considering the two functional reliability aspects, i.e., adequacy and the dynamic security or operating reliability.
Adequacy assessment becomes more complex due to the variability introduced by renewable energy generation. The traditionally used reserve margin only considers projected peak demand and would be inadequate since it does not consider an evaluation of off-peak conditions that could also be critical due to the variable renewable generation. Therefore, in order to address the impact of variable renewable generation, a probabilistic evaluation that studies all hours of a year based on statistical characteristics is a necessity to identify the adequacy risks. On the other hand, the system dynamic behavior is also changing. Converter interfaced generation resources have different dynamic characteristics from the conventional synchronous units and inherently do not participate in grid regulation functions such as frequency control and voltage control that are vital to maintaining operating reliability. In order to evaluate these evolving grid characteristics, comprehensive reliability evaluation approaches that consider system stochasticity and evaluate both adequacy and dynamic security are important to identify potential system risks in this transforming environment.
Adequacy assessment becomes more complex due to the variability introduced by renewable energy generation. The traditionally used reserve margin only considers projected peak demand and would be inadequate since it does not consider an evaluation of off-peak conditions that could also be critical due to the variable renewable generation. Therefore, in order to address the impact of variable renewable generation, a probabilistic evaluation that studies all hours of a year based on statistical characteristics is a necessity to identify the adequacy risks. On the other hand, the system dynamic behavior is also changing. Converter interfaced generation resources have different dynamic characteristics from the conventional synchronous units and inherently do not participate in grid regulation functions such as frequency control and voltage control that are vital to maintaining operating reliability. In order to evaluate these evolving grid characteristics, comprehensive reliability evaluation approaches that consider system stochasticity and evaluate both adequacy and dynamic security are important to identify potential system risks in this transforming environment.
ContributorsWang, Yingying (Author) / Vittal, Vijay (Thesis advisor) / Khorsand, Mojdeh (Thesis advisor) / Heydt, Gerald (Committee member) / Ayyanar, Raja (Committee member) / Arizona State University (Publisher)
Created2020
Description
Pursuit of an informed approach to interpreting Frédéric Chopin’s music has been increasingly challenging in the twenty-first century. In the process of forming their unique voices, pianists turn to the sound recordings of some of the most notable pianistic figures in history. This document offers a detailed inspection of three revered recordings and, with the help of syntactic analysis, seeks an understanding of the extraordinary interpretational decisions of Alfred Cortot, Arthur Rubinstein and Dinu Lipatti. The examined works are Chopin’s Prelude in C Major, Op. 28, No. 1, and the Largo of the Sonata in B Minor, Op. 58. The analysis of the Prelude compares recorded performances of Alfred Cortot (ca. 1933-1934) and Arthur Rubinstein (ca. 1946) and explains how their vastly different interpretational choices can, through an analytical process, be traced to the harmonic and melodic implications of the score. Likewise, inspection of the Largo focuses on Dinu Lipatti’s performance (ca. 1947) and draws connections between his phrasing and critical characteristics of the movement. All three performances present exquisite examples of a style of expressive playing that seems to have fallen into disuse in the twenty-first century. This study contributes to a deeper understanding of the performing style of Cortot, Rubinstein, and Lipatti, and also seeks to show connections between score analysis and interpretational decisions.
ContributorsJovanovic, Isidora (Author) / Pagano, Caio (Thesis advisor) / Holbrook, Amy (Thesis advisor) / Ryan, Russell (Committee member) / Arizona State University (Publisher)
Created2020