The goal of this study is to assess conventional and low temperature dual fuel combustion in light- and heavy-duty multi-cylinder compression ignition engines in terms of combustion characterization, performance, and emissions. First, a light-duty compression ignition engine is converted to a dual fuel engine and instrumented for in-cylinder pressure measurements. The primary fuels, methane and propane, are each introduced into the system by means of fumigation before the turbocharger, ensuring the airuel composition is well-mixed. Experiments are performed at 2.5, 5, 7.5, and 10 bar BMEP at an engine speed of 1800 RPM. Heat release analyses reveal that the ignition delay and subsequent combustion processes are dependent on the primary fuel type and concentration, pilot quantity, and loading condition. At low load, diesel-ignited propane yields longer ignition delay periods than diesel-ignited methane, while at high load the reactivity of propane is more pronounced, leading to shorter ignition delays. At high load (BMEP = 10 bar), the rapid heat release associated with diesel-ignited propane appears to occur even before pilot injection, possibly indicating auto-ignition of the propane-air mixture. Next, a modern, heavy-duty compression ignition engine is commissioned with an open architecture controller and instrumented for in-cylinder pressure measurements. Initial diesel-ignited propane dual fuel experiments (fumigated before the turbocharger) at 1500 RPM reveal that the maximum percent energy substitution (PES) of propane is limited to 86, 60, 33, and 25 percent at 5, 10, 15, and 20 bar BMEP, respectively. Fueling strategy, injection strategy, exhaust gas recirculation (EGR) rate, and intake boost pressure are varied in order to maximize the PES of propane at 10 bar BMEP, which increases from 60 PES to 80 PES of propane. Finally, diesel-ignited propane dual fuel low temperature combustion (LTC) is implemented using early injection timings (50 DBTDC) at 5 bar BMEP. A sweep of injection timings from 10 DBTDC to 50 DBTDC reveals the transition from conventional to low temperature dual fuel combustion, indicated by ultra-low NOx and smoke emissions. Optimization of the dual fuel LTC concept yields less than 0.02 g/kW-hr NOx and 0.06 FSN smoke at 93 PES of propane.

Deep neural networks (DNNs), particularly convolutional neural networks (CNNs), have been widely applied in the many fields, such as computer vision, natural language processing, speech recognition and etc. Although DNNs achieve dramatic accuracy improvements in these real-world tasks, they require significant amounts of resources (e.g., memory, energy, storage, bandwidth and computation resources). This limits the application of these networks on resource-constrained systems, such as mobile and edge devices. A large body of literature has been proposed to addresses this problem from the perspective of compressing DNNs while preserving their performance. In this thesis, we focus on compressing deep CNNs based on vector quantization techniques. The first part of this thesis summarizes some basic concepts in machine learning and popular techniques on model compression, including pruning, quantization, low-rank factorization and knowledge distillation approaches. Our main interest is quantization techniques, which compress networks by reducing the precision of parameters. Full-precision weights, activations and even gradients in networks can be quantized to 16-bit floating point numbers, 8-bit integers, or even binary numbers. Despite a possible performance degradation, quantization can greatly reduce the model size while maintaining model accuracy. In the second part of this thesis, we propose a novel vector quantization approach, which we refer to as Vector Quantization with Learned Codebook, or VQLC, for CNNs. Rather than performing scalar quantization, we choose vector quantization that can simultaneously quantize multiple weights at once. Instead of taking a pretraining/clustering approach as in most works, in VQLC, the codebook for quantization are learned together with neural network training from scratch. For the forward pass, the traditional convolutional filters are replaced by the convex combinations of a set of learnable codewords. During inference, the compressed model will be represented by a small-sized codebook and a set of indices, resulting in a significant reduction of model size while preserving the network's performance. Lastly, we validate our approach by quantizing multiple modern CNNs on several popular image classification benchmarks and compare with state-of-the-art quantization techniques. Our experimental results show that VQLC demonstrates at least comparable and often superior performance to the existing schemes. In particular, VQLC demonstrates significant advantages over the existing approaches on wide networks at the high rate of compression.

We study hyperbolic structures on the compression body C with genus 2 positive boundary and genus 1 negative boundary. We consider individual hyperbolic structures as well as special regions in the space of all such hyperbolic structures. We use some properties of the boundary hyperbolic structures on C to establish an interesting property of cusp shapes of tunnel number one manifolds. This extends a result of Nimershiem in [26] to the class of tunnel number one manifolds. We also establish convergence results on the geometry of compression bodies. This extends the work of Ito in [13] from the punctured-torus case to the compression body case.

This thesis explores the potential of edge devices in three applications: indoor localization, urban traffic prediction, and multi-modal representation learning. For indoor localization, we propose a reliable data transmission network and robust data processing framework by visible light communications and machine learning to enhance the intelligence of smart buildings. The urban traffic prediction proposes a dynamic spatial and temporal origin-destination feature enhanced deep network with the graph convolutional network to collaboratively learn a low-dimensional representation for each region to predict in-traffic and out-traffic for every city region simultaneously. The multi-modal representation learning proposes using dynamic contexts to uniformly model visual and linguistic causalities, introducing a novel dynamic-contexts-based similarity metric that considers the correlation of potential causes and effects to measure the relevance among images. To enhance distributed training on edge devices, we introduced a new system called Distributed Artificial Intelligence Over-the-Air (AirDAI), which involves local training on raw data and sending trained outputs, such as model parameters, from local clients back to a central server for aggregation. To aid the development of AirDAI in wireless communication networks, we suggested a general system design and an associated simulator that can be tailored based on wireless channels and system-level configurations. We also conducted experiments to confirm the effectiveness and efficiency of the proposed system design and presented an analysis of the effects of wireless environments to facilitate future implementations and updates. This thesis proposes FedForest to address the communication and computation limitations in heterogeneous edge networks, which optimizes the global network by distilling knowledge from aggregated sub-networks. The sub-network sampling process is differentiable, and the model size is used as an additional constraint to extract a new sub-network for the subsequent local optimization process. FedForest significantly reduces server-to-client communication and local device computation costs compared to conventional algorithms while maintaining performance with the benchmark Top-K sparsification method. FedForest can accelerate the deployment of large-scale deep learning models on edge devices.

As the need for digital libraries, especially genealogical libraries, continues to rise, the need for efficient document image compression is becoming more and more apparent. In addition, because many digital library users access them from dial-up Internet connections, efficient strategies for compression and progressive transmission become essential to facilitate browsing operations. To meet this need, we developed a novel method for representing document images in a parametric form. Like other “hybrid" image compression operations, the Contour Encoded Compression and Transmission (CECAT) system first divides images into foreground and background layers. The emphasis of this thesis revolves around improving the compression of the bitonal foreground layer. The parametric vectorization approach put forth by the CECAT system compares favorably to current approaches to document image compression. Because many documents, specifically handwritten genealogical documents, contain a wide variety of shapes, fitting Bezier curves to connected component contours can provide better compression than current glyph library or other codebook compression methods. In addition to better compression, the CECAT system divides the image into layers and tiles that can be used as a progressive transmission strategy to support browsing operations.

As computer generated special effects play an increasingly integral role in the development of films and other media, simulating granular material continues to be a challenging and resource intensive process. Solutions tend to be pieced together in order to address the complex and different behaviors of granular flow. As such, these solutions tend to be brittle, overly specific, and unnatural. With the introduction of a holistic 3-phase granular simulation, we can now create a reliable and adaptable granular simulation.Our solution improves upon this hybrid solution by addressing the issue of particle flow and correcting interpenetration amongst particles while maintaining the efficiency of the overall simulation. We achieve this by projecting particles onto a 2D manifold and implementing density correction using a quadratic solver. Particle updates are projected back into 3D to spread the particles apart on each frame.Keywords:

Software memory compression has been a common practice among operating systems. Since then, prior works have explored hardware memory compression to reduce the load on the CPU by offloading memory compression to hardware. However, prior works on hardware memory compression cannot provide critical isolation in multi-tenant systems like cloud servers. Our evaluation of prior work (TMCC) shows that a tenant can be slowed down by more than 12x due to the lack of isolation. This work, Compressed Memory Management Unit (CMMU), prototypes hardware compression for multi-tenant systems. CMMU provides critical isolation for multi-tenant systems.First, CMMU allows OS to control individual tenants' usage of physical memory. Second, CMMU compresses a tenant's memory to an OS-specified physical usage target. Finally, CMMU notifies the OS to start swapping the memory to the storage if it fails to compress the memory to the target. We prototype CMMU with a real compression module on an FPGA board. CMMU runs with a Linux kernel modified to support CMMU. The prototype virtually expands the memory capacity to 4X. CMMU stably supports the modified Linux kernel with multiple tenants and applications. While achieving this, CMMU only requires several extra cycles of overhead besides the essential data structure accesses. ASIC synthesis results show CMMU fits within 0.00931mm2 of silicon and operates at 3GHz while consuming 36.90mW of power. It is a negligible cost to modern server systems. / Master of Science / Memory is a critical resource in computer systems. Memory compression is a common technique to save memory resources. Memory compression consumes the computing resource, traditionally supplied by the CPU. In other words, memory compression traditionally competes with applications for CPU computing power. The prior work, TMCC, provides a design to perform memory compression in ASIC hardware, therefore no longer competing for CPU computing power. However, TMCC provides no isolation in a multitenant system like a modern cloud server. This thesis prototypes a new design, Compressed Memory Management Unit (CMMU), providing isolation in hardware memory compression. This prototype can speed up applications by 12x compared to without the isolation, with a 4x expansion in virtual memory capacity. CMMU supports a modified Linux OS running stably. CMMU also runs at high clock speed and offers little overhead in latency, silicon chip area, and power

No description available.

No description available.

No description available.