This thesis will discuss the role of government in venture capital market and illustrate the national difference of that through a comparative study of government expenditure supporting policies of venture capital investment in the USA, Canada and China. Firstly, the prototype programs designed by Small Business Administration (SBA) of the United States and Canadian policies by the Business Development Bank of Canada (BDC) will be discussed. Secondly, the current Chinese government expenditure supporting policies will then be studied in the context of the Chinese venture capital market’s unique political, economical and legislative background. Ultimately, potential improvements in the future expenditure incentive programs in China will be explored.

This thesis will discuss the role of government in venture capital market and illustrate the national difference of that through a comparative study of government expenditure supporting policies of venture capital investment in the USA, Canada and China. Firstly, the prototype programs designed by Small Business Administration (SBA) of the United States and Canadian policies by the Business Development Bank of Canada (BDC) will be discussed. Secondly, the current Chinese government expenditure supporting policies will then be studied in the context of the Chinese venture capital market’s unique political, economical and legislative background. Ultimately, potential improvements in the future expenditure incentive programs in China will be explored.

Les méthodes de réseau de tenseurs (TN) ont été un ingrédient essentiel des progrès de la physique de la matière condensée et ont récemment suscité l'intérêt de la communauté de l'apprentissage automatique pour leur capacité à représenter de manière compacte des objets de très grande dimension. Les méthodes TN peuvent par exemple être utilisées pour apprendre efficacement des modèles linéaires dans des espaces de caractéristiques exponentiellement grands [1]. Dans ce manuscrit, nous dérivons des limites supérieures et inférieures sur la VC-dimension et la pseudo-dimension d'une grande classe de Modèles TN pour la classification, la régression et la complétion . Nos bornes supérieures sont valables pour les modèles linéaires paramétrés par structures TN arbitraires, et nous dérivons des limites inférieures pour les modèles de décomposition tensorielle courants (CP, Tensor Train, Tensor Ring et Tucker) montrant l'étroitesse de notre borne supérieure générale. Ces résultats sont utilisés pour dériver une borne de généralisation qui peut être appliquée à la classification avec des matrices de faible rang ainsi qu'à des classificateurs linéaires basés sur l'un des modèles de décomposition tensorielle couramment utilisés. En corollaire de nos résultats, nous obtenons une borne sur la VC-dimension du classificateur basé sur le matrix product state introduit dans [1] en fonction de la dimension de liaison (i.e. rang de train tensoriel), qui répond à un problème ouvert répertorié par Cirac, Garre-Rubio et Pérez-García [2]. / Tensor network (TN) methods have been a key ingredient of advances in condensed matter physics and have recently sparked interest in the machine learning community for their ability to compactly represent very high-dimensional objects. TN methods can for example be used to efficiently learn linear models in exponentially large feature spaces [1]. In this manuscript, we derive upper and lower bounds on the VC-dimension and pseudo-dimension of a large class of TN models for classification, regression and completion. Our upper bounds hold for linear models parameterized by arbitrary TN structures, and we derive lower bounds for common tensor decomposition models (CP, Tensor Train, Tensor Ring and Tucker) showing the tightness of our general upper bound. These results are used to derive a generalization bound which can be applied to classification with low-rank matrices as well as linear classifiers based on any of the commonly used tensor decomposition models. As a corollary of our results, we obtain a bound on the VC-dimension of the matrix product state classifier introduced in [1] as a function of the so-called bond dimension (i.e. tensor train rank), which answers an open problem listed by Cirac, Garre-Rubio and Pérez-García [2].

Fundraising with venture capitalists can remain a largely mysterious process. In a world shrouded with non-disclosure agreements, the entrepreneurs are often unaware of the common practices of deal terms and are unable to benchmark their term sheets with respect to those given to others. Inherent conflicts of interest in the split of the financial returns, liquidation, and control of the company lead the venture capitalists to structure the deals which benefit their interests at cost to the interests of the entrepreneurs. This dissertation identifies and characterizes the term sheet structures used by venture capitalists today and establishes their frequency. This information can be used by entrepreneurs to benchmark their term sheets and by venture capitalists to evaluate their investment strategies.

In 1998, it was proved by Ben-David and Litman that a concept space has a sample compression scheme of size $d$ if and only if every finite subspace has a sample compression scheme of size $d$. In the compactness theorem, measurability of the hypotheses of the created sample compression scheme is not guaranteed; at the same time measurability of the hypotheses is a necessary condition for learnability. In this thesis we discuss when a sample compression scheme, created from compression schemes on finite subspaces via the compactness theorem, have measurable hypotheses. We show that if $X$ is a standard Borel space with a $d$-maximum and universally separable concept class $\m{C}$, then $(X,\CC)$ has a sample compression scheme of size $d$ with universally Borel measurable hypotheses. Additionally we introduce a new variant of compression scheme called a copy sample compression scheme.

Student Number : 0109917P - PhD thesis - School of Process and Materials Engineering - Faculty of Engineering and the Built Environment / The aim of this project was to characterise new WC-10VC-Co powders, and to deposit WC-10VC-Co thermal spray coatings from these powders for characterisation and testing in adhesion, wear and corrosion tests. Throughout the project, the new powders and coatings were compared to commercial WC-Co powders of the same binder content and commercial WC-Co thermal spray coatings. All the powders i.e WC-10VC-Co and WC-Co powders, were produced by agglomeration (by spray drying) and sintering and characaterised by determining the sizes and size distributions of the powders' particles, the morphology, the flowability and the phase composition and grain size and size distribution of carbide grains. The vanadium carbide in the WC-10VC-Co powders occurred in the solution as the double carbide (V,W)C and the carbides present in the WC-10VC-Co powders were WC and (V,W)C. None of the starting VC was left in the powders. Coatings were deposited using high pressure high velocity oxy-fuel (HP/HVOF) spraying systems, and characterized by determining the microstructures, the phase compositions and the carbide grain sizes, as had been done for the powders. Three types of tests were done on the coatings: adhesion tests, (according to standard SNECMA 14 -008); dry abrasion, wet abrasion and slurry erosion tests; and corrosion tests, in synthetic mione water. Thermal spraying lead to some WC decarburization to W2C and eta phase, and to the formation of amorphous binder. The W2C grains from the WC decarburization formed in the amorphous binder matrix of coatings. All the coatinge were porous, but the new WC-10VC-Co coatings were more porous than the commercial Wc-Co coatings because the spray parameters had only been optimized for the WC-Co coatings. The carbide grains decreased in size by as much as 50% during decomposition. Evidence suggested that the WC grains in the coatings were subjected to different residual stresses that in the powders, probably due to the formation of the amorphous binder. Vanadium carbide in the Wc-10VC-Co coatings occurred as (V,W)C, just as in the powders, with as distribution that was reasonably homogeneous. The apparent hardness of the new Wc-10VC-Co coatings was slightly lower than that of WC-Co coatings of the same cobalt content, due to their higher porosity. The adhesion of the new Wc-10VC-Co coatings was as good as that of the Wc-Co coatings. The dry and wet abrasion resistance of the new Wc-10VC-Co coatings was better that for the Wc-Co coatings of equal Co wt%, on account of the Wc-10VC-Co coatings having a lower binder volume fraction, finer carbide grains, and (V,W)C grains. The (V,W)C grains are harder than WC grains and apparently slowed down the overall abrasion rate. In slurry erosion, the best performance of the Wc-10VC-Co coatings was as good as that of the commercial WC-Co coatings at equal cobalt mass content, due to the higher porosity of the Wc-10VC-Co coatings, apparent faster erosion of the harder but brittle (V,W)C grains, and, from what evidence appreared to suggest, generally slightly poorer erosion resistance of the fine WC grains under the test conditions used. Polishing the slurry erosion test specimens reduced mass losses in slurry erosion by factor of up to 10 compared to the unpolished specimens, and led to better erosion resistance of the WC-10VC-Co coating compared to the WC-12Co coating. The results of the tests done to investigate the corrosion properties of the coatings were conclusive. This is because the effects of cleaning procedures on mass loss after immersion corrosion were not explored, and it appeared, for some coatings, that the corrosion mechanisms in immersion corrosion could not be reproduced in electrochemical testing.

With a wide range of potential applications, the machine type communication (MTC) is gaining a tremendous interest among mobile network operators, system designers, MTC specialist companies, and research institutes. The idea of having electronic devices and systems automatically connected to each other without human intervention is one of the most significant objectives for future wireless communications. Low data rate transmission and the requirement for low energy consumption are two typical characteristics for MTC applications. In terms of supporting low cots MTC devices, industrial standards will be more efficient if designers can re-use many features of existing radio access technologies. This will yield a cost effective solution to support MTC in future communication systems. This thesis investigates efficient MTC waveform and receiver designs for superior signal transmission quality with low operational costs. In terms of the downlink receiver design, this thesis proposes a novel virtual carrier (VC) receiver system for MTC receivers, which aims to reduce the maximum bandwidth to improve the data processing efficiency and cost-efficiency by using analogue filters to extract only sub-carriers of interest. For the VC receiver systems, we thus reduce the sampling rate in order to reduce the number of subsequent processing operations, which significantly reduces the analogue-to-digital converter (ADC) cost and power consumption while providing high signal to interference noise ratio (SINR) and low bit to error rate (BER) to support low data rate MTC devices. Our theoretical equations account for the interference effect of aliasing on the sub-carrier location, and this helps the system designer to evaluate what kind of filters and receiver sampling rate can be used to balance the energy cost and detection performance. In terms of the uplink waveform design, considering the enhanced number of MTC devices in the future communication systems, i.e. the fifth generation (5G) communications, the same tight synchronisation as used in today appears not to be cost-effective or even possible. Synchronisation signals, which aim to provide a perfect time or frequency synchronisation in the current fourth generation (4G) communication systems (known as the long-term evolution, LTE), is much more costly for low data rate MTC transmissions. The system bandwidth will be significantly reduced if a base station tries to synchronise all received signals among hundreds or thousands MTC devices in one transmission time period. In terms of relaxing the synchronisation requirements, this thesis compares and analyses the side-lobe reduction performance for several candidate multi-carrier waveforms to avoid these problems. We also propose the infinite impulse response universal filtered multi-carrier (UFMC) system and the overlap and add UFMC system, which significantly reduce the processing complexity compared with the state of the art UFMC techniques. This thesis derives closed-form expressions for the interference caused by time offsets between adjacent unsynchronised MTC users. Our analytical equations can be used in both simple and complex time-offset transmission scenarios, and enable the system designer to evaluate the SINR, the theoretical Shannon capacity and the BER performance.

In 1998, it was proved by Ben-David and Litman that a concept space has a sample compression scheme of size $d$ if and only if every finite subspace has a sample compression scheme of size $d$. In the compactness theorem, measurability of the hypotheses of the created sample compression scheme is not guaranteed; at the same time measurability of the hypotheses is a necessary condition for learnability. In this thesis we discuss when a sample compression scheme, created from compression schemes on finite subspaces via the compactness theorem, have measurable hypotheses. We show that if $X$ is a standard Borel space with a $d$-maximum and universally separable concept class $\m{C}$, then $(X,\CC)$ has a sample compression scheme of size $d$ with universally Borel measurable hypotheses. Additionally we introduce a new variant of compression scheme called a copy sample compression scheme.

Video codec compresses the input video source to reduce storage and transmission bandwidth requirements while maintaining the quality. It is an essential technology for applications, to name a few such as digital television, DVD-Video, mobile TV, videoconferencing and internet video streaming. There are different video codecs used in the industry today and understanding their operation to target certain video applications is the key to optimization. The latest advanced video codec standards have become of great importance in multimedia industries which provide cost-effective encoding and decoding of video and contribute for high compression and efficiency. Currently, H.264 AVC, AVS, and DIRAC are used in the industry to compress video. H.264 codec standard developed by the ITU-T Video Coding Experts Group (VCEG) together with the ISO/IEC Moving Picture Experts Group (MPEG). Audio-video coding standard (AVS) is a working group of audio and video coding standard in China. VC-2, also known as Dirac Pro developed by BBC, is a royalty free technology that anyone can use and has been standardized through the SMPTE as VC-2. H.264 AVC, Dirac Pro, Dirac and AVS-P2 are dedicated to High Definition Video, while AVS-P7 is to mobile video. Out of many standards, this work performs a comparative analysis for the H.264 AVC, DIRAC PRO/SMPTE-VC-2 and AVS-P7 standards in low bitrate region and high bitrate region. Bitrate control and constant QP are the methods which are employed for analysis. Evaluation parameters like Compression Ratio, PSNR and SSIM are used for quality comparison. Depending on target application and available bitrate, order of performance is mentioned to show the preferred codec.

The current trend of digital convergence creates the need for the video encoder and decoder system, known as codec in short, that should support multiple video standards on a single platform. In a modern video codec, quantization is a key unit used for video compression. In this thesis, a generalized quantization algorithm and hardware implementation is presented to compute quantized coefficient for six different video codecs including the new developing codec High Efficiency Video Coding (HEVC). HEVC, successor to H.264/MPEG-4 AVC, aims to substantially improve coding efficiency compared to AVC High Profile. The thesis presents a high performance circuit shared architecture that can perform the quantization operation for HEVC, H.264/AVC, AVS, VC-1, MPEG- 2/4 and Motion JPEG (MJPEG). Since HEVC is still in drafting stage, the architecture was designed in such a way that any final changes can be accommodated into the design. The proposed quantizer architecture is completely division free as the division operation is replaced by multiplication, shift and addition operations. The design was implemented on FPGA and later synthesized in CMOS 0.18 μm technology. The results show that the proposed design satisfies the requirement of all codecs with a maximum decoding capability of 60 fps at 187.3 MHz for Xilinx Virtex4 LX60 FPGA of a 1080p HD video. The scheme is also suitable for low-cost implementation in modern multi-codec systems.