This research investigates three solutions to overcoming portability and scalability concerns in the Evolutionary Hardware (EHW) field. Firstly, the study explores if the V-FPGA—a new, portable Virtual-Reconfigurable-Circuit architecture—is a practical and viable evolution platform. Secondly, the research looks into two possible ways of making EHW systems more scalable: by optimising the system’s genetic algorithm; and by decomposing the solution circuit into smaller, evolvable sub-circuits or modules. GA optimisation is done is by: omitting a canonical GA’s crossover operator (i.e. by using an algorithm); applying evolution constraints; and optimising the fitness function. The circuit decomposition is done in order to demonstrate modular evolution. Three two-bit multiplier circuits and two sub-circuits of a simple, but real-world control circuit are evolved. The results show that the evolved multiplier circuits, when compared to a conventional multiplier, are either equal or more efficient. All the evolved circuits improve two of the four critical paths, and all are unique. Thus, it is experimentally shown that the V-FPGA is a viable hardware-platform on which hardware evolution can be implemented; and how hardware evolution is able to synthesise novel, optimised versions of conventional circuits. By comparing the and canonical GAs, the results verify that optimised GAs can find solutions quicker, and with fewer attempts. Part of the optimisation also includes a comprehensive critical-path analysis, where the findings show that the identification of dependent critical paths is vital in enhancing a GA’s efficiency. Finally, by demonstrating the modular evolution of a finite-state machine’s control circuit, it is found that although the control circuit as a whole makes use of more than double the available hardware resources on the V-FPGA and is therefore not evolvable, the evolution of each state’s sub-circuit is possible. Thus, modular evolution is shown to be a successful tool when dealing with scalability.

Neuronal networks, derived from mouse embryonic frontal cortex (FC) tissue grown on microelectrode arrays, were used to investigate effects of gentamicin pretreatment on pharmacological response to the L-type calcium channel blocker, verapamil. Gentamicin is a broad spectrum antibiotic used to control bacterial contamination in cell culture. The addition of gentamicin directly to medium affects the pharmacological and morphological properties of the cells in culture. A reproducible dose response curve to verapamil from untreated cultures was established and the mean EC50 was calculated to be 1.5 ± 0.5 μM (n=10). 40 μM bicuculline was added to some cell cultures to stabilize activity and verapamil dose response curves were performed in presence of bicuculline, EC50 1.4 ± 0.1 μM (n=9). Statistical analysis showed no significant difference in verapamil EC50s values obtained in presence of bicuculline and hence the data was combined and a standard verapamil EC50 was calculated as 1.4 ± 0.13 μM (n=19). This EC50 was then used to compare verapamil EC50s obtained from neuronal cell cultures with chronic and acute exposures to gentamicin. FC cultures (21- 38 days old) were found to be stable in presence of 2300 μM gentamicin. The recommended concentration of gentamicin for contamination control is 5uL /1 ml medium (108 μM). At this concentration, the verapamil EC50 shifted from 1.4 ± 0.13 μM to 0.9 ± 0.2 μM. Given the limited data points and only two complete CRCs, statistical comparison was not feasible. However, there is a definite trend that shows sensitization of cells to verapamil in presence of gentamicin. The cultures exposed to 108 μM gentamicin for 5 days after seeding showed loss of adhesion and no data could be collected for pharmacological analysis. To conclude, acute gentamicin exposure of neuronal cell cultures causes increased sensitivity to verapamil and chronic or long term exposure to gentamicin may cause loss of adhesion of the cell culture by affecting the glial growth. The effect of chronic exposure to gentamicin on pharmacological responses to verapamil remains inconclusive.

Ingeniero Civil Eléctrico / La síntesis de haces resuelve el problema de generar patrones de radiación más complejos que los que se puede generar con antenas individuales o arreglos de estas. Para ello se manipula la fase y la amplitud de la alimentación de cada elemento del arreglo, lo que permite focalizar la emisión o recepción de la potencia en ciertos lugares del espacio. Tanto en radioastronomía, como en las telecomunicaciones, esta tecnología da la posibilidad de observar el espacio de manera selectiva y/o en múltiples direcciones de manera simultanea, lo cual permite disminuir tiempos de barrido (para el caso de radiotelescopios) o descongestionar canales creando enlaces directivos. Este trabajo presenta la implementación de un sintetizador digital de múltiples haces, utilizando un arreglo planar de de 4x4 antenas tipo parche, capaz de generar tres patrones patrones directivos, cuyo procesamiento se realiza íntegramente en una FPGA (Field Programmable Gate Array). El arreglo está sintonizado para operar en 5.81 GHz, con un ancho de banda de 40 MHz. La señal es convertida a banda base por un conjunto de tarjetas de electrónica analógica. Se han realizado dos pruebas experimentales para verificar el funcionamiento del sistema: una emulando condiciones ideales de excitación, de modo de eliminar interferencias ajenas al experimento, y otra situando el arreglo en un extremo de una habitación vacía, siendo iluminado por una fuente puntual en el otro extremo. Los resultados permiten concluir que la síntesis de haces es satisfactoria, generando haces cuyo lóbulo principal es de 20º de ancho, con la capacidad rotar la dirección en que apuntan en todo el hemisferio visible del arreglo.

Over the next decade, multi-antenna radios, including phased array and multiple-input-multiple-output (MIMO) radios, are expected to play an essential role in the next-generation of wireless networks. Phased arrays can reject spatial interferences and provide coherent beamforming gain, and MIMO technology promises to significantly enhance the system performance in the coverage, capacity, and user data rate through the beamforming or diversity/capacity gain which can substantially increase the range in wireless links, that are challenged from the transmitter (TX) power handling, receiver (RX) noise perspectives and a multi-path environment. Furthermore, the multi-user MIMO (MU-MIMO) can simultaneously serve multiple users which is vital for femtocell base stations and access points (AP). Full-duplex (FD) wireless, namely simultaneous transmission and reception at the same frequency, is an emerging technology that has gained attention due to its potential to double the data throughput, as well as provide other benefits in the higher layers such as better spectral efficiency, reducing network and feedback signaling delays, and resolving hidden-node problems to avoid collisions. However, several challenges remain in the quest for the high-performance integrated FD radios. Transmitter power handling remains an open problem, particularly in FD radios that integrate a shared antenna interface. Secondly, FD operation must be achieved across antenna VSWR variations and a changing EM environment. Finally, FD must be extended to multi-antenna radios, including phased array and multi-input multi-output (MIMO) radios, as over the next decade, they are expected to play an essential role in the next generation of wireless networks. Multi-antenna FD operation, however, is challenged not only by the self-interference (SI) from each TX to its own RX but also cross-talk SI (CT-SI) between antennas. In this dissertation, first, a full-duplex phased array circulator-RX (circ.-RX) is proposed that achieves self-interference cancellation (SIC) through repurposing beamforming degrees of freedom (DoF) on TX and RX. Then, an FD MIMO circ.-RX is proposed that achieves SI and CT-SI cancellation (CT-SIC) through passive RF and shared-delay baseband (BB) canceller that addresses challenges associated with FD MIMO operation. Wireless radios at millimeter-wave (mm-wave) frequencies enable the high-speed link for portable devices due to the wide-band spectrum available. Large-scale arrays are required to compensate for high path loss to form an mm-wave link. Mm-wave MIMO systems with digitization enable virtual arrays for radar, digital beamforming (DBF) for high mobility scenarios and spatial multiplexing. To preserve MIMO information, the received signal from each element in MIMO RX should be transported to ADC/DSP IC for DBF, and vice versa on the TX side. A large-scale array can be formed by tiling multiple mm-wave IC front-ends, and thus, a single-wire interface is desired between DSP IC and mm-wave ICs to reduce board routing complexity. Per-element digitization poses the challenge of handling high data-rate I/O in large-scale tiled MIMO mm-wave arrays. SERializer – DESerializer (SERDES) is traditionally being used as a high-speed link in computing systems and networks. However, SERDES results in a large area and power consumption. In this dissertation, a 60~GHz 4-element MIMO TX with a single-wire interface is presented that de-multiplexes the baseband signal of all elements and LO reference that are frequency-domain multiplexed on a single-wire coax cable.

The aim of this work is primarily to validate the advanced techniques for treatment planning and dosimetric verification for modern megavoltage x-ray radiotherapy. With the advent of modern radiotherapy techniques, there is a great need for assuring quality of the radiation dose distributions generated by the advanced intensity modulated treatments (IMRT/VMAT). This is typically accomplished by the assessment of the treatment plan quality at the planning stage and then verification of the dose distributions through measurements on the phantoms or independent dose calculations prior to the actual delivery of these plans to patients. The major focus of this work is to clinically evaluate the modern 2D and 3D dose verification techniques. The measurement-based dosimetry systems investigated were ArcCHECK/3DVH and SRS MapCHECK. AcrCHECK/3DVH system uses the measurement-guided dose reconstruction algorithm to correct the predicted dose in the patient dataset. The system was intended for VMAT/IMRT QA. SRS MapCHECK was investigated for SRS treatments. The independent dose calculation system was DoseCHECK which employed a GPU-accelerated convolution-superposition of algorithm for 3D dose reconstruction on the patient dataset. Next, a hybrid dose verification system (PerFRACTION) was evaluated, which takes input from both the treatment planning system and the linac EPID and produces a measurement-guided 3D dose distribution for comparison with the plan. This system was investigated for potential QA applications to a modern, efficient SRS technique, involving simultaneously treating multiple targets with a single isocenter. The performance of all dosimetry systems was validated against well-characterized independent dosimeters, such as ion chamber, film and scintillator detectors, or 3D arrays (Delta4), using stringent dose comparison criteria to test their limits for the intended clinical applications. For the initial plan quality evaluation of a novel tool (Feasibility DVH) was investigated. This tool a priori estimates best achievable dose volume histograms for a specific patient, based on the basic physics properties of the megavoltage x-rays, thus helping the planners to guide their efforts. All studied dosimetry systems showed an excellent agreement of the average gamma (a mathematical combination of DD and DTA) passing rates >98% for most of the plans. The 3% DD/2mm DTA criteria were used for extracranial plans and 3%/1mm for intracranial SRS plans. As dictated by the logic of the application, the comparisons were made against TPS calculations, a bi-planar array, or film measurements. Similarly the average percent point dose errors <2% were observed against the ion chambers or film. In the rare instances when the deviations were larger, intuitive explanations were provided, based on either the physics of the plans or inhomogeneous patient anatomy and resulting algorithm limitations. Feasibility DVH was shown to reliably predict the best possible organ sparing for clinical head-and-neck VMAT plans. Overall the investigated dosimetry systems were found reliable and feasible for their intended clinical use.

The integral image representation of an image is important for a large number of modern image processing algorithms. Integral image representations can reduce computation and increase the operating speed of certain algorithms, improving real-time performance. Due to increasing demand for real-time image processing performance, an integral image architecture capable of accelerating the calculation based on the amount of available resources is presented. Use of the proposed accelerator allows for subsequent stages of a design to have data sooner and execute in parallel. It is shown here how, with some additional resources used in the Field Programmable Gate Array (FPGA), a speed increase is obtained by using a one-dimensional Systolic Array (SA) approach. Additionally, extra guidelines are given for further research in this area.

No description available.

Clustering is the process of finding underlying group structure in data. As the scale of data collection continues to grow, this “big data” phenomenon results in more complex data structures. These data structures are not always compatible with traditional clustering methods, making their use problematic. This thesis presents methodology for analyzing samples of four-way and higher data, examples of these more complex data types. These data structures consist of samples of continuous data arranged in multidimensional arrays. A large emphasis is placed on clustering this data using mixture models that leverage tensor-variate distributions to model the data. Parameter estimation for all these methods are based on the expectation-maximization algorithm. Both simulated and real data are used for illustration. / Thesis / Doctor of Science (PhD)

No description available.

No description available.