Este trabalho trata de pesquisa de soluções para a síntese física de circuitos integrados menos susceptíveis aos efeitos de variabilidade decorrentes do uso de tecnologias de fabricação com dimensões nanométricas. Também apresenta a pesquisa e o desenvolvimento de uma ferramenta para a geração de leiautes regulares denominada R-CAT. A regularidade geométrica é explorada pela repetição de padrões básicos de leiaute ao longo de uma matriz. A regularidade é apontada como uma das melhores alternativas para lidar com os atuais problemas de fabricação em tecnologias submicrônicas. Projetos regulares são menos suscetíveis aos problemas de litografia, aumentam o yield e diminuem o tempo gasto em re-projeto. Além disso, circuitos regulares apresentam maior previsibilidade de resultados de potência, atraso e yield, principalmente pelo fato das células estarem pré-caracterizadas. A ferramenta desenvolvida visa o trabalho com dois tipos de síntese física para leiautes regulares, produzindo circuitos integrados personalizáveis por todas as máscaras ou circuitos personalizáveis por algumas máscaras. O principal objetivo deste gerador é a facilidade de conversão e adaptação dependendo da abordagem de matriz escolhida. Isso facilitará a comparação entre diferentes alternativas de matrizes, a adoção de blocos lógicos diversos e de novas tecnologias. O gerador de leiautes R-CAT identifica células adjacentes com conexões em comum entre elas e realiza a conexão entre essas células em metal 1, reduzindo o número de conexões a ser realizado pelo roteador em até 10%. A ferramenta R-CAT está inserida em um fluxo de projeto e depende do método de síntese lógica adotado. Duas ferramentas de síntese lógica foram utilizadas: SIS e OrBDDs, oferecendo duas linhas de projeto: a primeira priorizando a área e a segunda priorizando timing e interconexões curtas. Ambas respeitando a mesma regularidade geométrica imposta pela matriz. Os resultados obtidos demonstram que as matrizes SIS ocupam 53% menos área do que a estratégia orBDD e reduzem o wire length em 30%. Uma área menor é obtida devido ao fato da ferramenta SIS gerar descrições com a metade de células lógicas e nets. Entretanto, as matrizes R-CAT OrBDD apresentam menor wire length médio, menor fan-out (redução de 15%), menor delay e maior roteabilidade. As sínteses OrBDD apresentam poucas nets não roteadas sem a inserção de trilhas extras. Além disso, as matrizes R-CAT atingiram resultados até 40% menores em wire length e reduções de área de até 46% em relação às matrizes MARTELO. / This work presents a research for physical synthesis of integrated circuits, which are less susceptible to the effects of variability observed in fabrication technologies using nanometers scale. Moreover, it presents a CAD tool developed to generate regular layouts, which is called R-CAT. The geometric regularity is achieved using basic patterns repeated along one matrix structure. Regularity is pointed like one of the best alternatives to deal with submicron technologies issues. Regular designs are less susceptible to lithographic problems, improve the yield and decrease the time to re-spin. Furthermore, regular circuits improve predictability of power consumption, timing and yield results, because the cells are pre-characterized. The developed tool focuses on two types of physical synthesis for regular layouts, producing either integrated circuit customized using all masks or integrated circuits customized using some masks. The main goal is the facility of conversion and adaptation depending on the chosen matrix approach. This will make easier the comparison of different matrix approaches, besides the adoption of several logic blocks and new technologies. R-CAT layout generator identifies adjacent cells that are placed in a same row and have common connections between them. In this case, the generator can make these connections in Metal 1. This technique reduces the number of connections to be done by the router. The experiments showed that this technique is able to reduce about 10% the number of connections to be done. This tool is inserted into a design flow and it is dependent of the logic synthesis methodology adopted. Two logical syntheses tools were used in the flow: SIS and OrBDDs. R-CAT SIS and R-CAT orBDD Matrices were generated for a set of circuits. The use of R-CAT tool with SIS and orBDD logical synthesis offers two design lines: the first one highlights area and the second one emphasize timing and short connections. Both of them respect the same geometric regularity. The results demonstrate that SIS matrices present 53% less area than orBDD approach and reduce the wire length by 30%. The area reduction is achieved because the SIS tool generates descriptions with the half of logic cells and nets. Nevertheless, the R-CAT orBDD matrices decreased the medium wire length, reduced the fan-out in 15%, reduced the delay and improved the routability. orBDD synthesis presents few non-routed nets without extra tracks insertion. Moreover, the R-CAT matrices obtained about 40% better results in wire length and they reduced area in 46% when compared to MARTELO matrices.

abstract: Cyber-physical systems and hard real-time systems have strict timing constraints that specify deadlines until which tasks must finish their execution. Missing a deadline can cause unexpected outcome or endanger human lives in safety-critical applications, such as automotive or aeronautical systems. It is, therefore, of utmost importance to obtain and optimize a safe upper bound of each task’s execution time or the worst-case execution time (WCET), to guarantee the absence of any missed deadline. Unfortunately, conventional microarchitectural components, such as caches and branch predictors, are only optimized for average-case performance and often make WCET analysis complicated and pessimistic. Caches especially have a large impact on the worst-case performance due to expensive off- chip memory accesses involved in cache miss handling. In this regard, software-controlled scratchpad memories (SPMs) have become a promising alternative to caches. An SPM is a raw SRAM, controlled only by executing data movement instructions explicitly at runtime, and such explicit control facilitates static analyses to obtain safe and tight upper bounds of WCETs. SPM management techniques, used in compilers targeting an SPM-based processor, determine how to use a given SPM space by deciding where to insert data movement instructions and what operations to perform at those program locations. This dissertation presents several management techniques for program code and stack data, which aim to optimize the WCETs of a given program. The proposed code management techniques include optimal allocation algorithms and a polynomial-time heuristic for allocating functions to the SPM space, with or without the use of abstraction of SPM regions, and a heuristic for splitting functions into smaller partitions. The proposed stack data management technique, on the other hand, finds an optimal set of program locations to evict and restore stack frames to avoid stack overflows, when the call stack resides in a size-limited SPM. In the evaluation, the WCETs of various benchmarks including real-world automotive applications are statically calculated for SPMs and caches in several different memory configurations. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2017

Este trabalho trata de pesquisa de soluções para a síntese física de circuitos integrados menos susceptíveis aos efeitos de variabilidade decorrentes do uso de tecnologias de fabricação com dimensões nanométricas. Também apresenta a pesquisa e o desenvolvimento de uma ferramenta para a geração de leiautes regulares denominada R-CAT. A regularidade geométrica é explorada pela repetição de padrões básicos de leiaute ao longo de uma matriz. A regularidade é apontada como uma das melhores alternativas para lidar com os atuais problemas de fabricação em tecnologias submicrônicas. Projetos regulares são menos suscetíveis aos problemas de litografia, aumentam o yield e diminuem o tempo gasto em re-projeto. Além disso, circuitos regulares apresentam maior previsibilidade de resultados de potência, atraso e yield, principalmente pelo fato das células estarem pré-caracterizadas. A ferramenta desenvolvida visa o trabalho com dois tipos de síntese física para leiautes regulares, produzindo circuitos integrados personalizáveis por todas as máscaras ou circuitos personalizáveis por algumas máscaras. O principal objetivo deste gerador é a facilidade de conversão e adaptação dependendo da abordagem de matriz escolhida. Isso facilitará a comparação entre diferentes alternativas de matrizes, a adoção de blocos lógicos diversos e de novas tecnologias. O gerador de leiautes R-CAT identifica células adjacentes com conexões em comum entre elas e realiza a conexão entre essas células em metal 1, reduzindo o número de conexões a ser realizado pelo roteador em até 10%. A ferramenta R-CAT está inserida em um fluxo de projeto e depende do método de síntese lógica adotado. Duas ferramentas de síntese lógica foram utilizadas: SIS e OrBDDs, oferecendo duas linhas de projeto: a primeira priorizando a área e a segunda priorizando timing e interconexões curtas. Ambas respeitando a mesma regularidade geométrica imposta pela matriz. Os resultados obtidos demonstram que as matrizes SIS ocupam 53% menos área do que a estratégia orBDD e reduzem o wire length em 30%. Uma área menor é obtida devido ao fato da ferramenta SIS gerar descrições com a metade de células lógicas e nets. Entretanto, as matrizes R-CAT OrBDD apresentam menor wire length médio, menor fan-out (redução de 15%), menor delay e maior roteabilidade. As sínteses OrBDD apresentam poucas nets não roteadas sem a inserção de trilhas extras. Além disso, as matrizes R-CAT atingiram resultados até 40% menores em wire length e reduções de área de até 46% em relação às matrizes MARTELO. / This work presents a research for physical synthesis of integrated circuits, which are less susceptible to the effects of variability observed in fabrication technologies using nanometers scale. Moreover, it presents a CAD tool developed to generate regular layouts, which is called R-CAT. The geometric regularity is achieved using basic patterns repeated along one matrix structure. Regularity is pointed like one of the best alternatives to deal with submicron technologies issues. Regular designs are less susceptible to lithographic problems, improve the yield and decrease the time to re-spin. Furthermore, regular circuits improve predictability of power consumption, timing and yield results, because the cells are pre-characterized. The developed tool focuses on two types of physical synthesis for regular layouts, producing either integrated circuit customized using all masks or integrated circuits customized using some masks. The main goal is the facility of conversion and adaptation depending on the chosen matrix approach. This will make easier the comparison of different matrix approaches, besides the adoption of several logic blocks and new technologies. R-CAT layout generator identifies adjacent cells that are placed in a same row and have common connections between them. In this case, the generator can make these connections in Metal 1. This technique reduces the number of connections to be done by the router. The experiments showed that this technique is able to reduce about 10% the number of connections to be done. This tool is inserted into a design flow and it is dependent of the logic synthesis methodology adopted. Two logical syntheses tools were used in the flow: SIS and OrBDDs. R-CAT SIS and R-CAT orBDD Matrices were generated for a set of circuits. The use of R-CAT tool with SIS and orBDD logical synthesis offers two design lines: the first one highlights area and the second one emphasize timing and short connections. Both of them respect the same geometric regularity. The results demonstrate that SIS matrices present 53% less area than orBDD approach and reduce the wire length by 30%. The area reduction is achieved because the SIS tool generates descriptions with the half of logic cells and nets. Nevertheless, the R-CAT orBDD matrices decreased the medium wire length, reduced the fan-out in 15%, reduced the delay and improved the routability. orBDD synthesis presents few non-routed nets without extra tracks insertion. Moreover, the R-CAT matrices obtained about 40% better results in wire length and they reduced area in 46% when compared to MARTELO matrices.

Predictive anisotropy is a phenomenon referring to unequal predictability of surface wind components in different directions. This study addresses the question of whether predictive anisotropy resulting from statistical prediction is influenced by physical factors or by types of regression methods (linear vs nonlinear) used to construct the statistical prediction. A systematic study of statistical predictability of surface wind components at 2109 land stations across the globe is carried out. The results show that predictive anisotropy is a common characteristic for both linear and nonlinear statistical prediction, which suggests that the type of regression method is not a major influential factor. Both strong predictive anisotropy and poor predictability are more likely to be associated with wind components characterized by relatively weak and non-Gaussian variability and in areas characterized by surface heterogeneity. An idealized mathematical model is developed separating predictive signal and noise between large-scale (predictable) and local (unpredictable) contributions to the variability of surface wind, such that small signal-to-noise ratio (SNR) corresponds to low and anisotropic predictability associated with non-Gaussian local variability. The comparison of observed and simulated statistical predictability by Regional Climate models (RCM) and reanalysis in the Northern Hemisphere indicates that small-scale processes that cannot be captured well by RCMs contribute to poor predictability and strong predictive anisotropy in observations. A second idealized mathematical model shows that spatial variability in specifically the minimum directional predictability, resulting from local processes, is the major contributor to predictive anisotropy. / Graduate

This thesis includes three essays on topics related to the predictability of market returns. I investigate i) the predictability of market returns from an adjusted version of cay ratio (cayadj), ii) the explanatory power of a conditional version of the consumption-CAPM which uses predictor variables to scale the pricing kernel, and iii) whether information about future market returns can be extracted from a large set of commodity data. The first essay studies the predictive ability of cayadj . In Campbell and Mankiw (1989), the consumption-wealth ratio is represented as a linear function of expected market returns and consumption growth. Lettau and Ludvigson (2001) build their study on Campbell and Mankiw (1989) and estimate the ratio cay as a proxy for the consumption-wealth ratio, assuming that the fluctuation in expected consumption growth is constant. I argue that the variation in expected consumption growth should be taken into consideration and propose adjusting the cay ratio by the estimates of expected consumption growth. After making the adjustment, I find that the predictabilities of market returns, particularly at annual, bi-annual, and tri-annual horizons, are greatly improved. The significant predictive ability of cayadj still holds in out-of-sample forecasts. The second essay examines the performance of a conditional version of the consumption-CAPM, where conditioning variables are used to scale the pricing kernel. I find that incorporating the conditioning information into the standard consumption-CAPM greatly improves the performance in asset pricing tests, particularly when using cayadj as the conditioning variable. Moreover, the performance of conditional consumption-CAPM is as good as the ultimate consumption risk model (Parker and Julliard, 2005) which measures the consumption risk over several quarters. Further tests show that the factors of conditional consumption-CAPM drive out the consumption risk measured over several quarters. The third essay evaluates the ability of lagged commodity returns to forecast market returns. In order to exploit the predictive information from a relatively large amount of commodity returns, I apply the partial-least-squares (PLS) method pioneered by Kelly and Pruitt (2013). I find that the commodity returns measured over previous twelve months show strong predictive power in monthly and three-month forecasts, in-sample and out-of-sample. The findings are robust to controlling for risk factors such as momentum, Fama-French three factors and industry returns previously identified to be significant predictors of market returns (Hong, Torous and Valkanov, 2007).

O objetivo deste estudo foi analisar o controle postural diante da previsibilidade da carga elevada através de ação manual e da demanda atencional promovida pelo acréscimo de um alvo espacial na tarefa. Quinze sujeitos do sexo masculino (idade média 22,8+3,5 anos) realizaram a tarefa de elevação de uma caixa por meio da preensão manual, com massa variável (1, 3 e 5kg) nas condições previsível ou imprevisível. A tarefa também era executada em uma condição de maior demanda atencional promovida pela colocação de um alvo frontal, que deveria ser atingido por um facho de luz laser, cuja fonte estava fixada na caixa elevada pelo sujeito. Foram analisados o comportamento do centro de pressão (CP) e a cinemática das principais articulações e do movimento da caixa. Os resultados indicaram que a imprevisibilidade sobre a massa da caixa aumentou a oscilação do CP. Aumento da oscilação do CP também foi observado na situação de redução de massa, mesmo quando os sujeitos tinham conhecimento da mudança. A situação de pontaria em alvo induziu redução (a) do deslocamento do CP, (b) velocidade de elevação da caixa, e (c) amplitude de movimento do tornozelo e quadril. Estes resultados indicam que a previsibilidade sobre a ação de forças externas no corpo, demanda espacial da tarefa manual, e organização/feedback do movimento da tentativa anterior interagem para afetar o controle postural / The aim of this study was to analyze postural control as a function of predictability of a load lifted with the hands, and demand of spatial accuracy in the task. Fifteen males (mean age: 22.8 3.5 years) lifted boxes with variable masses (1, 3 and 5 kg), in conditions of predictable or unpredictable load. The task was performed either under lower spatial accuracy demand or aiming at a spatial target. Displacement of center of pressure (CP) and kinematics of the main joints/box motion were analyzed to access postural perturbation induced by the lifting task. Results showed that unpredictability about mass of the box enhanced CP displacement. Enhanced CP displacement was also observed in situations of decreased mass regarding the previous trial, even when participants were aware about the change. The pointing task induced decreased (a) CP displacement, (b) box lifting velocity, and (c) ankle and hip motion amplitude. These results indicate that predictability about external forces acting on the body, spatial demand of manual tasks, and movement organization/feedback of the previous trial interact to affect postural control

In this dissertation, I study corporate activities, and their predictive abilities of market returns. The first chapter examines the determinants of industry merger waves. We propose a continuous merger activity variable (MAV) as an alternative to discrete industry merger waves. We find that the ranking of MAV within a quarter is associated with strong patterns in before and after industry returns and operating performance. During 1985-2015, bucket 1 containing industries with lowest MAV rank earns alpha of 0.30% per month higher than bucket 12 containing industries with highest MAV rank. The second chapter examines the predictive ability of many corporate activities, including mergers and acquisitions, insider trading, share repurchases, etc. Using machine learning approaches, we find that an aggregate index of corporate activities has substantial predictive power of future market returns both in- and out-of-sample, and yields much greater economic gain for a mean-variance investor. We further find that the predictive ability of the corporate index stems from its information content about future cash flows and expected corporate investments and that the corporate index performs particularly well for stocks with greater information asymmetry. The third chapter examines the relationship between firm valuation and takeover activity, using the European debt crisis as a laboratory. The European debt crisis in mid-2011 caused a wide-spread redemption of money market mutual funds (MMFs) with high exposure to European borrowers. The shock to a group of MMFs induced a reduction in the volume of repurchase agreements (repos) that the funds were engaged in. The resulting decline in the amount of capital available to support equity positions (furnished as collateral by the repo counterparties) potentially lead to downward price pressure on these stocks. I find that the temporary shock in firm valuation triggers more opportunistic takeover bids as suggested by prior studies, but it does not increase the takeover success rate. Overall, my findings present a more complicated picture to the relationship between market valuation and takeover activities than suggested by prior studies.

As the decline in Arctic sea ice extent makes this region more accessible, the need is increasing for effective seasonal sea ice forecasting to facilitate operational planning. Recently, coupled global climate models (CGCMs) have been used to address the need for effective sea ice forecasting on seasonal time scales. This thesis assesses the operational utility of the Canadian Seasonal to Interannual Prediction System (CanSIPS) for seasonal sea ice forecasting. This assessment consists of two separate studies. The first uses hindcasting to analyze the skill of two versions of CanSIPS, as well as an intermediate version, on the pan-Arctic as well as regional scales. This approach allows for an overall assessment of the system's skill in addition to providing insight with regards to the features in each version which improved that skill. This study finds that the use of a new initialization procedure for sea ice concentration and thickness improved forecast skill on the pan-Arctic scale as well as in the Central Arctic, Barents Sea, Laptev Sea, and Sea of Okhotsk. This study also shows that the substitution of one of the constituent models in the system improved forecast skill on the pan-Arctic scale as well as in the GIN, Barents, Kara, East Siberian, Chukchi, Bering, and Beaufort Seas. Overall, the new version of CanSIPS was found to be generally more skillful than previous versions. The second study conducts a potential predictability experiment on CanCM4, the constituent CGCM common to all versions of CanSIPS considered in this study. This study follows the methodology introduced by \cite{Bushuk2018} which allows for a more complete assessment of the dependency of potential predictability on initialization month than previous studies and for comparisons to be made between potential predictability and operational skill. This analysis is again done on both the pan-Arctic and regional scale. The findings of this experiment show that CanCM4 has relatively low potential predictability relative to other models and explains results previously presented in a multi-model study by \cite{Day2016}. Further, the characteristics of CanCM4's potential predictability share similarities with other models including greater predictability at longer lead times for winter target months than summer target months, greater predictability in the Atlantic sector than the Pacific sector, and the presence of the spring predictability barrier on the pan-Arctic scale as well as in several regions. The comparison of operational skill to potential predictability provides a general overview of the ``skill gap" which may be closed with improvements in initialization procedures and model physics. This comparison does, however, come with some caveats due to differences in the statistical characteristics of the perfect model and the climate system it represents. Together, the operational skill assessment of different versions of CanSIPS and the potential predictability experiment conducted on one of its constituent models, CanCM4, demonstrate that while room for improvement exists, the recent development of this forecast system has clearly increased its operational utility as a seasonal sea ice forecasting tool. / Graduate

<p>In Chapter 1, I construct a new proxy for Tobin's q that incorporates the replacement cost of patent capital. This proxy, PI (physical plus intangible) q, explains up to 64\% more variation in investment than other proxies for q. Furthermore, investment is more sensitive to PI q than to other proxies for q. Although investment is predicted more accurately by, and is more sensitive to, PI q, controlling for PI q leads to relatively higher, not lower, cash flow coefficients. All results are stronger in subsamples with more patent capital. Overall, using PI q strengthens the historically weak investment-q relation.</p> <p><br></p> <p>Chapter 2 includes Noah Stoffman and M. Deniz Yavuz as co-authors, and in this chapter, we find that small innovators (i.e., small, innovative firms) earn higher returns than small non-innovators for up to five years. We find no such innovative premium among large firms. A battery of tests shows that our results are explained by risk, not investor underreaction. Small innovators are especially risky because they focus more on risky product innovation and rely more on organization capital that amplifies their systematic risk. In addition, small innovators contribute significantly to the size premium. Overall, small innovators have a higher cost of equity, which potentially explains why they rely heavily on internal capital.</p>

We examine the predictability of expected stock returns across horizons using machine learning. We use neural networks, and gradient boosted regression trees on the U.S. and international equity datasets. We find that predictabil- ity of returns using neural networks models decreases with longer forecasting horizon. We also document the profitability of long-short portfolios, which were created using predictions of cumulative returns at various horizons, be- fore and after accounting for transaction costs. There is a trade-off between higher transaction costs connected to frequent rebalancing and greater returns on shorter horizons. However, we show that increasing the forecasting hori- zon while matching the rebalancing period increases risk-adjusted returns after transaction cost for the U.S. We combine predictions of expected returns at multiple horizons using double-sorting and buy/hold spread, a turnover reduc- ing strategy. Using double sorts significantly increases profitability on the U.S. sample. Buy/hold spread portfolios have better risk-adjusted profitability in the U.S. JEL Classification G11, G12, G15, C55 Keywords Machine learning, asset pricing, horizon pre- dictability, anomalies Title Multi-horizon equity returns predictability via machine learning