Search

Many recent research efforts have exploited data sparsity for the acceleration of convolutional neural network (CNN) inferences. However, the effects of data transfer between main memory and the CNN accelerator have been largely overlooked. In this work, the authors propose a CNN acceleration technique that leverages hardware/software co‐design and exploits the sparsity in input feature maps (IFMs). On the software side, the authors' technique employs a novel lossless compression scheme for IFMs, which are sent to the hardware accelerator via direct memory access. On the hardware side, the authors' technique uses a CNN inference accelerator that performs convolutional layer operations with their compressed data format. With several design optimization techniques, the authors have implemented their technique in a field‐programmable gate array (FPGA) system‐on‐chip platform and evaluated their technique for six different convolutional layers in SqueezeNet. Results reveal that the authors' technique improves the performance by 1.1×–22.6× while reducing energy consumption by 47.7%–97.4% as compared to the CPU‐based execution. Furthermore, results indicate that the IFM size and transfer latency are reduced by 34.0%–85.2% and 4.4%–75.7%, respectively, compared to the case without data compression. In addition, the authors' hardware accelerator shows better performance per hardware resource with less than or comparable power consumption to the state‐of‐the‐art FPGA‐based designs.

The subject of the research is the financial mechanism of the modern venture investment market under conditions of uncertainty and the requirements of accelerated innovative development. The study aims to analyze the impact of the pandemic on the sustainability of the institutional behavior of venture capital investment in the global and Russian markets. The authors apply such scientific methods as sampling, grouping, comparison, analogy, analysis, generalization, systemic-structural approach to assessing the development of the subject of research. The paper analyses the role of venture investment in the successful promotion of innovations using the example of the world’s leading companies in 2016–2021. Successful examples of venture investment are presented. The study investigates the impact of the pandemic and lockdown on the level of volatility, composition and dynamics of venture capital investments in 2020– 2021. The authors conclude that the pandemic has an impact on the sectoral redistribution of investments in favor of venture investments in medicine, biotechnology, the sector of information and communication technologies and solutions for business, education, healthcare, on market polarization (covid-negative, covid-positive, covid-neutral), on reducing the role of state funds and Russian accelerators, while increasing the interest and scale of participation of private institutions (foreign investors, business angels, syndicated angel investment), to increase of the volume of the Russian venture capital market by increasing the average ticket size despite the reduction in their number, to strengthen the practice reproduction of their own ecosystems by corporate institutions, to popularize venture capital investment deals in terms of mergers and acquisitions. Prospects for researching the venture capital market are associated with the constant attention to high-tech covid-positive projects, with the growth of transactions in the early stages, due to the increase in the number of professional communities, with the pinpoint development of specific tasks of an investor or corporation by accelerators, with venture investment of companies with the subsequent entry of the project into their own ecosystem, with the improvement of the state’s restrictive measures aimed at a clear institutionalization of professional activities and increasing the financial stability of participants in the investment and innovation markets, with the stimulation of mechanisms for regional innovative development to attract capital to high-risk projects in the regions.

Purpose: Business accelerators have rapidly emerged as prominent players in the entrepreneurial ecosystem. A key strategic decision in designing acceleration programs is whether to customize or standardize the new venture development program (Cohen et al., 2019). Recognizing a trade-off between customization and standardization, the paper presents a multistage acceleration model aiming to harvest benefits of standardization while keeping several advantages found in tailor-made acceleration programs.

With the determination of the number of new innovative startups businesses, the situation is even more complicated, not only in our country, but also in most countries of the world. In the proposed work, the author made an attempt to implement one of the possible approaches to introducing transparency – an assessment of each of the segments of the startup-launches ecosystem separately. The paper gives the results of study on the evaluation the launch indicators of new startups in the Russian venture market. The article identifies the segments of the venture ecosystem that are directly involved in the launch of new startups. The study obtained an estimate of the number of startups, small and medium-sized companies as well as entrepreneurial startups. To characterize the number of new startups, the author introduces the term performance of segments of the ecosystem of innovative startups. This indicator allows you to use the quantitative characteristics of individual segments and compare them with each other. To obtain quantitative values of newly launched startups, the author conducted a selective telephone survey of representatives of these segments of the ecosystem and collected official data published on the Internet. The author carried out the processing of some data on the basis of the statistics of median values of the estimate of the number of new startups that started sales during 2019. The author collected data for 2020 from secondary sources. The study showed that 4,817 new startups with signs of innovation were launched in Russia in 2020, and an increase of 26 % compared to the previous year. The estimates obtained do not contradict with the published data on the state of the venture industry during the pandemic.

The Open Radio Access Network (RAN) concept has been gaining wide acceptance in recent network architectures, including 5G New Radio (NR) network deployments. Current Open RAN radio implementation efforts, aim at integrating several white-box hardware elements and executing digital processing on open-source software. When building such a software-based, 5G Open RAN platform, challenges include achieving real-time execution times for demanding computational blocks of the 5G NR physical layer processing, such as Low Density Parity Check (LDPC) decoding. In this context, having already identified both the capabilities as well as the challenges that Field-programmable Gate Arrays (FPGAs) offer for accelerating LDPC, we present our novel LDPC FPGA accelerator system. In this paper, we contribute the implementation details of our FPGA accelerator design as well as the process of integrating the accelerator with OpenAirInterface (OAI), the basis for our 5G NR platform. For the first time in the literature, we show an FPGA-based LDPC accelerator fully integrated with a complete software platform, that is able to achieve more than $1.6Gbps$ decoding throughput and up to 13 times faster execution times compared to single core software implementations. Finally, in our results, we show that LDPC encoding is more challenging to accelerate due to lower computational complexity.

Convolutional Neural Networks (CNNs) have been shown to be very useful in image recognition and other Artificial Intelligence (AI) applications, however, at the expense of intensive computation requirement. To address the challenge of overwhelming calculation requirements, researchers have proposed various network pruning techniques. But, due to the irregular sparse patterns, unstructured sparse networks are difficult to compute efficiently on either Graphic processing units (GPUs) or Field Programmable Gate Arrays (FPGAs). In this paper, we propose a software/hardware co-optimized Reconfigurable Sparse convolutional Neural Network accelerator design (RSNN) on FPGAs. A novel sparse convolution dataflow is proposed with simpler control logic than existing mux-based selection logic. To balance the computation load on different Processing Units (PUs), we propose a software-based load-balance aware pruning technique as well as a kernel merging method. Experimental results show that RSNN is $2.41\times - 7.91\times $ better on Digital Signal Processor (DSP) efficiency than previous dense CNN FPGA accelerators, and $1.23\times - 2.93\times $ better than previous sparse CNN FPGA accelerators.

The increasing significance of technology in daily lives led to the need for the development of convenient methods of human-computer interaction (HCI). Given that the existing HCI approaches exhibit various limitations, hand gesture recognition-based HCI may serve as a more intuitive mode of human-machine interaction in many situations. In addition, the system has to be deployable on low-power devices for applicability in broadly defined Internet of Things (IoT) and smart home solutions. Recent advances exhibit the potential of deep learning models for gesture classification, whereas they are still limited to high-performance hardware. Embedded neural network accelerators are constrained in terms of available memory, central processing unit (CPU) clock speed, graphics processing unit (GPU) performance, and a number of supported operations. The aforementioned problems are addressed in this paper by namely two approaches - simplifying the signal processing pipeline to avoid recurrent structures and efficient topological design. This paper employs an intuitive scheme allowing for the generation of the data in the compressed form from the sequence of range-Doppler images (RDI). Thus, it allows for the design of a neural classifier avoiding the usage of recurrent layers. The proposed framework has been optimized for Intel® Neural Compute Stick 2 (Intel® NCS 2), at the same time achieving promising classification accuracy of 97.57%. To confirm the robustness of the proposed algorithm, five independent persons have been involved in the algorithm testing process.

This paper aimed to identify what are the strategic benefits that the social incubation process offers for social enterprise. We conducted a case study with a Brazilian social incubator. Social enterprises that participated in the social incubation program were the analysis units. Among the eight strategic benefits found, three of them have never been previously indicated in the studied literature, therefore being a novelty herein highlighted to improve social enterprise's commercial and social missions. By understanding these benefits, social entrepreneurs can make better decisions about joining social incubation. Besides, managers of social incubators can understand the limitations inherent to the social incubation herein stressed, and, from this, improve the programs. Hence, this paper advances knowledge in the social entrepreneurship field. Este artículo tuvo como objetivo identificar cuáles son los beneficios estratégicos que ofrece el proceso de incubación social para las empresas sociales. Se realizó un estudio de caso con una incubadora social brasileña. Las unidades de análisis fueron constituidas por empresas sociales que participaron en el programa de incubación. Se identificaron ocho beneficios estratégicos, tres de los cuales no están en la literatura analizada, siendo una novedad deste artículo. Al comprender estos beneficios, los emprendedores sociales pueden tomar mejores decisiones sobre adentrarse en una incubación social. Además, los gerentes de incubadoras sociales pueden comprender las limitaciones de la incubación social destacadas aquí para mejorar los programas. Así, este artículo avanza el conocimiento en el campo del emprendimiento social. Este artigo objetivou identificar quais são os benefícios estratégicos que o processo de incubação social oferece para os empreendimentos sociais. Um estudo de caso foi realizado com uma incubadora social brasileira. As unidades de análise foram constituídas por empreendimentos sociais que participaram do programa de incubação. Oito benefícios estratégicos foram identificados, três deles não constam na literatura analisada, sendo uma novidade aqui destacada para melhorar as missões comerciais/sociais dos empreendimentos sociais. Entendendo esses benefícios, os empreendedores sociais podem tomar melhores decisões sobre o ingresso na incubação social. Gestores de incubadoras sociais podem entender as limitações inerentes à incubação social aqui destacadas e, a partir delas, aprimorar os programas. Portanto, este artigo avança o conhecimento no campo do empreendedorismo social.

Designing resource-efficient deep neural networks (DNNs) is a challenging task due to the enormous diversity of applications as well as their time-consuming design, training, optimization, and evaluation cycles, especially the resource-constrained embedded systems. To address these challenges, we propose a novel DNN design framework called accuracy-and-performance-aware neural architecture search (APNAS), which can generate DNNs efficiently, as it does not require hardware devices or simulators while searching for optimized DNN model configurations that offer both inference accuracy and high execution performance. In addition, to accelerate the process of DNN generation, APNAS is built on a weight sharing and reinforcement learning-based exploration methodology, which is composed of a recurrent neural network controller as its core to generate sample DNN configurations. The reward in reinforcement learning is formulated as a configurable function to consider the sample DNNs' accuracy and cycle count required to run on a target hardware architecture. To further expedite the DNN generation process, we devise analytical models for cycle count estimation instead of running millions of DNN configurations on real hardware. We demonstrate that these analytical models are highly accurate and provide cycle count estimates identical to those of a cycle-accurate hardware simulator. Experiments that involve quantitatively varying hardware constraints demonstrate that APNAS requires only 0.55 graphics processing unit (GPU) days on a single Nvidia GTX 1080Ti GPU to generate DNNs that offer an average of 53% fewer cycles with negligible accuracy degradation (on average 3%) for image classification compared to state-of-the-art techniques.

The energy efficiency of accelerating hundreds of MB-large deep neural networks (DNNs) in a mobile environment is less than that of a server-class big chip accelerator because of the limited power budget, silicon area, and smaller buffer size of static random access memory associated with mobile systems. To address this challenge and provide powerful computing capability for processing large DNN models in power/resource-limited mobile systems, we propose McDRAM v2, which is a novel in-dynamic random access memory (DRAM) systolic array accelerator architecture. McDRAM v2 makes the best use of large in-DRAM bandwidths for accelerating various DNN applications. It can handle large DNN models without off-chip memory accesses, in a fast and efficient manner, by exposing the large DRAM capacity and large in-DRAM bandwidth directly to an input systolic array of a processing element matrix. Additionally, it maximizes data reuse using a systolic multiply-accumulate (MAC) structure. The proposed architecture maximizes the utilization of large-scale MAC units by judiciously exploiting the DRAM's internal bus and buffer structure. An evaluation of large DNN models in the fields of image classification, natural language processing, and recommendation systems shows that it achieves 1.7 times tera operations per second (TOPS), 3.7 times TOPS/watt, and 8.6 times TOPS/mm2 improvements over a state-of-the-art mobile graphics processing unit accelerator, and 4.1 times better energy efficiency over a state-of-the-art server-class accelerator. Moreover, it incurs a minimal overhead, i.e., a 9.7% increase in area, and uses less than 4.4 W of peak operating power.

Recurrent neural networks such as Long Short-Term Memories (LSTMs) learn temporal dependencies by keeping an internal state, making them ideal for time-series problems such as speech recognition. However, the output-to-input feedback creates distinctive memory bandwidth and scalability challenges in designing accelerators for RNNs. We present Muntaniala, an RNN accelerator architecture for LSTM inference with a silicon-measured energy-efficiency of 3.25 TOP/s/W and performance of 30.53 GOP/s in UMC 65nm technology. The scalable design of Muntaniala allows running large RNN models by combining multiple tiles in a systolic array. We keep all parameters stationary on every die in the array, drastically reducing the I/O communication to only loading new features and sharing partial results with other dies. For quantifying the overall system power, including I/O power, we built Vau da Muntanialas, to the best of our knowledge, the first demonstration of a systolic multi-chip-on-PCB array of RNN accelerator. Our multi-die prototype performs LSTM inference with 192 hidden states in 330 μs with a total system power of 9.0 mW at 10 MHz consuming 2.95 μJ. Targeting the 8/16-bit quantization implemented in Muntaniala, we show a phoneme error rate (PER) drop of approximately 3% with respect to floating-point (FP) on a 3L-384NH-123NI LSTM network on the TIMIT dataset. ISSN:1549-8328 ISSN:1057-7122 ISSN:1558-0806

The numerical integration of the exchange–correlation (XC) potential is one of the primary computational bottlenecks in Gaussian basis set Kohn–Sham density functional theory (KS-DFT). To achieve optimal performance and accuracy, care must be taken in this numerical integration to preserve local sparsity as to allow for near linear weak scaling with system size. This leads to an integration scheme with several performance critical kernels which must be hand optimized for each architecture of interest. As the set of available accelerator hardware goes more diverse, a key challenge for developers of KS-DFT software is to maintain performance portability across a wide range of computational architectures. In this work, we examine a modular software design pattern which decouples the implementation details of performance critical kernels from the expression of high-level algorithmic workflows in a device-agnostic language such as C++; thus allowing for developers to target existing and emerging accelerator hardware within a single code base. We consider the efficacy of such a design pattern in the numerical integration of the XC potential by demonstrating its ability to achieve performance portability across a set of accelerator architectures which are representative of those on current and future U.S. Department of Energy Leadership Computing Facilities.

The next generation of the IceCube Neutrino Observatory, IceCube-Gen2, will constitute a much larger detector, increasing the rate of high-energy neutrinos. IceCube-Gen2 will address the long-standing questions about astrophysical accelerators. The experiment will also include a surface air-shower detector which will allow for measurements of cosmic rays in the energy region where a transition between Galactic and extragalactic accelerators is expected. As a baseline design for the surface detector, we consider a surface array above the optical in-ice array consisting of the same type of stations used for the IceTop enhancement, i.e., scintillation detectors and radio antennas. In order to better understand the capabilities of such an array, we performed simulations of its response to air showers, including both detector types. We will show the results of this simulation study and discuss the prospects for the surface array of IceCube-Gen2., Comment: Presented at the 37th International Cosmic Ray Conference (ICRC 2021). See arXiv:2107.06968 for all IceCube-Gen2 contributions

Start-up accelerators, a relatively new and competitive type of business incubators, are nowadays considered to be a highly effective way of providing venture support. Start-ups who have been supported through accelerators have an approximately 23% higher survival rate than other new businesses. This positive effect can be explained by the highly selective process accelerators apply when deciding on which venture projects to support. It comes as no surprise that understanding this process and the respective selection criteria is at the core of accelerator/incubator literature within entrepreneurship research. Existing research is however limited to the investigation of commercial accelerators which provide support to start-ups having an economic purpose only. Hence those academic findings cannot be simply extrapolated to accelerators supporting ventures combining economic goals with social and ecological purpose. Given the growing meaning of sustainability entrepreneurship and hence the increasing number of sustainability-oriented accelerators, the above limitation seems to be an important research gap. This paper addresses the above gap by investigating the selection processes and criteria of so-called impact accelerators focusing on the support of start-ups expected to create not only economic outcomes but also positive social and/or environmental impact. Building on existing accelerator literature, we qualitatively investigate the selection processes and criteria of nine European impact accelerators. By comparing our findings with existing research, we identify important differences between the selection approach of commercial and impact accelerators thus contributing to sustainability entrepreneurship research and practice.

Introduction. Relativistic, high intensity and small emittance electron bunches are the basis of a future linear collider and free electron laser projects. Drive beam generation in a wakefield structure employing for power extraction and acceleration low loss dielectrics like microwave ceramics, fused silica and Chemical Vapor Deposition (CVD) diamond were considered.Objective. We report here our experimental testing of a ceramic material with extremely low loss tangent at GHz frequency ranges allowing the realization of high efficiency wakefield acceleration. We also present Barium Strontium Titanium oxides (BST) ferroelectric material, which is a critical tuning element of the 400 MHz superconducting radiofrequency (RF) tuner developed and tested by the CERN/Euclid Techlabs collaboration. The materials discussed here also include quartz and CVD diamonds that are capable of supporting the high RF electric fields generated by electron beams or pulsed high power microwaves. These materials have been optimized or specially designed for accelerator applications.Materials and methods. The ceramic materials for accelerators, commonly used for the dielectric based accelerating structures, have to withstand high gradient accelerating fields, and prevent potential charging by electron beams. Correspondingly, the ceramic materials, fused silica and CVD diamond were tested with high power wakefield accelerating structures at Argonne Wakefield Accelerator of Argonne National Laboratory. Some of the presented here ceramic materials were tested at X-band 11.4 GHz magnicon high power source.Results. Low loss microwave ceramics, fused silica, and CVD diamonds have been considered as materials for dielectric based accelerating structures to study of the physical limitations encountered driving > 100 MV/m at microwave and ~ GV/m at THz frequencies in a dielectric based wakefield accelerator. Various ceramic compositions were high power and electron beam tested at X-band 11.4 GHz magnicon power source and Argonne Wakefield Accelerator correspondingly. Special attention was paid to the CVD diamond cylindrical Ka-band 35 GHz wakefield structure development. Finally, the dielectric based structure tuning was demonstrated by varying the permittivity of the BST ferroelectric layer by temperature changes and by applying an external direct current electric field across the ferroelectric. This allows us to control the effective dielectric constant of the composite system and therefore, to control the structure frequency during operation. The same type of ferroelectric material was used for the Ferroelectric Fast Reactive tuner (FE-FRT) development. In a world first, CERN has tested the prototype FE-FRT with a superconducting cavity, and frequency tuning has been successfully demonstrated.Conclusion. Recent results on the development and experimental testing of advanced dielectric materials for accelerator applications are presented. Low loss microwave ceramics, quartz and CVD diamond are considered. We presented our experimental results on wakefield generation in microwave frequency ranges with the dielectric based accelerating structures. Special attention was paid to the experimental results on high power testing at X-band of the externally powered dielectric based components. Finally, we present here first experimental demonstration of ferroelectric tunable microwave ceramic for accelerator application, which includes both tunable dielectric wakefield accelerating structure and ferroelectric based fast high power tuner for superconducting cavities. The experimental results presented here are critical for the advanced dielectric wakefield accelerating structures and other components development intended for the future linear collider projects.

The behavior of the shotcrete linings during the tunnel construction is complex due to the variability of its mechanical characteristics during the curing time. After the installation of the support structure, the lining is loaded along with the excavation face. A new calculation procedure involving two analytical methods, i.e. the convergence-confinement method and hyperstatic reaction method, have been developed. By means of these two methods, it is possible to assess the evolution of the stress state in the lining, and therefore, also of the safety factor with respect to the failure in compression of the shotcrete. Due to the analysis of the safety factor evolution over time, it is possible to correctly design the lining, to choose the type of sprayed concrete and to define the maximum admissible advance rate of the excavation face, in order not to critically load the lining. In the following paper, after having shown the definition of the safety factor, a parametric analysis is performed, in order to investigate the evolution of the safety factor of the lining for two different rock types, three different shotcrete types and two tunnel advance rates have been considered.

Through this study, we seek to understand the impact of the use of the living lab approach on product and business development in an eHealth accelerator. In the case accelerator, 20 startups developed innovative products atop the European FIWARE Future Internet technology platform. The novel design element of the case accelerator was the use of the living lab approach that was included for the purpose of engaging end users in the development and testing of new product prototypes. Our main result is that the living lab approach provided added value to participating companies and resulted in changes in their product development and marketing strategies. Overall, the case accelerator and the use of the living lab approach had a significant impact on the development, growth, and market success of the companies. Based on the results of the case accelerator, we propose the generic accelerator model presented by Pauwels and co-authors in 2016 to be extended with a new design element, the living lab approach.

The convolutional neural network (CNN) has been widely applied in computer vision applications. Due to the intensive computation, the general central processing unit (CPU) processors are not efficient to meet the real-time requirement. Various hardware accelerators based on the application specific integrated circuit (ASIC) and field programmable gate array (FPGA) have been designed to accelerate CNN models during the inference phase. The data flow architecture has been extensively adopted because of the high parallelism. However, given the continual development in the computer vision field, CNN models have become increasingly diverse. Thus, CNN accelerators based on data flow architectures face an emerging challenge to maintain high throughput while coping with various CNN models. In this paper, we design a software-defined architecture to solve this. The goal of this study is to make the hardware change as the application changes to achieve high flexibility and high performance. In our proposed architecture, all the parts can be software-defined to cope with different CNN models. A flexible software-defined processing element (PE) array is designed to compute different weight filter sizes. In addition, a software-defined data reuse technique based on two ideal reuse cases is proposed to ensure that all the parameters need to be loaded only once during the computing phase. To support this reuse technique, we also propose the software-defined on-chip buffer so that the weight and image buffers share one dynamic buffer. By using the sparsity property of the input feature map, the full-connected (FC) layer is accelerated. About 88% of the FC weight parameters can be skipped when loading the VGG-16 model. Finally, we implemented this software-defined accelerator on the FPGA. Compared to the other FPGA based accelerators, our proposed accelerator can preserve high performance while maintaining flexibility.

The aim of the article is to determine the main guidelines for the movement of startups in Ukraine as well as to study and assess the existing experience of working with international organizations whose activities are aimed at developing youth entrepreneurship and national higher education institutions (universities) to improve and deepen the cooperation. The essence of a startup is disclosed in the article, various definitions of the term “startup” are considered, the main trends in the development of the startup movement in Ukraine in general and youth entrepreneurship in particular are highlighted. Furthermore, the authoritative Global Startup Ecosystem Ranking is analyzed, and its evaluation criteria are given; the ranking of countries in terms of the number of startups and the Global Innovation Index is considered. Famous Ukrainian startups are studied, and their brief description is provided for a clearer understanding. The foreign experience in building ecosystem infrastructure, in particular, the experience of Singapore and Beijing, is reviewed. In addition, the national ecosystem for supporting youth start-ups in Ukraine and its main areas are analyzed. Moreover, the most important ecosystem projects are described, and their essence is determined. Examples of effective interaction between Ukrainian universities and international business incubators are given, in particular, the activities of the Innovative Ecosystem Sikorsky Challenge, a program aimed to support innovations and startups, at universities, and the positive experience of cooperation in the development of youth startups between the network of international academic business incubators YEP and the Faculty of Economics of V.N. Karazin Kharkiv National University is presented. The main activities of the business incubator, the history of its functioning are studied, and the opportunities that students receive while working with YEP in each of the presented directions are described.

Purpose This study aims to explore Accelerators and their practices in sustaining start-ups within their innovative programs for these companies based on the resource-based perspective. Moreover, with an ever-increasing demand for Accelerators amongst start-up companies, this study also demonstrates the importance of Accelerators, as it pertains to new venture creation. Design/methodology/approach This research uses an exploratory case study approach to examine a comparative view of leading Accelerator companies in the USA and Korea based on resource support. Findings The results of this study show that there are a number of differences between Accelerators of the two countries in terms of the resources they support for early-stage start-ups. The findings also show some similarities. However, in Korea, the Accelerator landscape is limited, where mentorship, resources and investments are not readily accessible, resulting in low success rates for Korean start-up companies. These limitations have had a negative trickle-down effect when providing entrepreneurs with strong access to resources and investors, which highly affects the success rates of early-stage start-ups. Practical implications In terms of the resource-based theory, this study contributes to the growth of early start-ups by emphasizing the role of the accelerator and suggesting the extent and impact that entrepreneurs have access to resources and investors. Originality/value With significant growth in start-ups around the world, the necessity for start-up funding and mentorship has increased drastically. Start-up companies need various types of assets, systems, knowledge and information to achieve their goals. In Accelerators, start-ups receive all the aforementioned resources while also improving their entrepreneurial skills. Start-up companies have many options in seeking investors who support both tangible and intangible resources to boost growth. While there is a wealth of information on traditional funding methods, there are few studies that shed light on the role of Accelerators from the resource-based point of view.

We present the initial design studies and specifications for an accelerator and conveyor system to irradiate collagen samples, modifying properties such as the putrescibility and mechanical behaviours in a paradigm shift from existing, widely used technology. We show the integrated design requirements for a magnetic rastering scheme to move the beam position in order to ensure a uniform dose distribution over the full surface of the hide and discuss its dependence on factors such as the size of the hide, the beam current and conveyor speed. We also present initial energy deposition studies using beam particle interaction simulation program G4beamline, in order to determine the numerical beam parameters and angle of incidence needed to ensure a uniform depth-dose distribution throughout the hide thickness.

Rutherford backscattering spectrometry (RBS) in channeling regimes (RBS/C), as an ion beam analysis method performed on a Tandetron 6MV accelerator, generally gives precise information about the structure of crystalline samples by combining RBS signals in the random and aligned configurations. This paper presents details about the design and implementation of tailored RBS/C measurements (coarse and fine) and data evaluation application developed in MATLAB for in situ accelerator control system ARGUS, delivered by High Voltage Engineering Europa BV (HVEE). Additionally, we examined two different ways of stepping during the measurement to reduce the possible inaccuracies related with goniometer’s backslash affecting the evaluation of spectra. Verification experiment was carried out using a 2-MeV 4He+-beam directed on a Si (100) substrate. The channeling effect is seen as channeling dips of a lower signal in an otherwise rather homogeneous plane. Implemented application significantly facilitates the RBS/C measurement and analysis of the experiments, and also extends the ion beam analysis portfolio of Advanced Technologies Research Institute. Finally, software is ready-to-use for any Tandetron based ion beam facility with the ARGUS software for accelerator control.

(Early Access) This paper introduces the first general reuse-centric accelerator for CNN inferences. Unlike prior work that exploits similarities only across consecutive video frames, general reuse-centric accelerator is able to discover similarities among arbitrary patches within an image or across independent images, and translate them into computation time and energy savings. Experiments show that the accelerator complements both prior software-based CNN and various CNN hardware accelerators, producing up to 14.96X speedups for similarity discovery, up to 3.33X speedups for a convolutional layer.

Entrepreneurship education (EE) is a means to create economic activity and is of importance in the development of rural regions. Entrepreneurship education is generally conducted by three different methods: (1) teaching students about entrepreneurial themes, (2) developing the capabilities of becoming a successful entrepreneur for enterprise and (3) developing capabilities through enterprise by which persons learn through experience and discovery. Our research question is, How does education through entrepreneurship fit as a way to meet the needs of start-ups in a rural context? We follow an entrepreneurial case programme in our study by conducting observations of and interviews with 26 persons involved in the programme. Our findings are that participants consider the programme as an important contribution to the local/regional entrepreneurial ecosystem. However, it is also described as a short-lived ecosystem, from which the participants mostly benefit while on the programme and to a lesser degree afterwards. Sustainability plays a part in attracting participants but to a lesser degree forms the educational content and evaluation parameters. Our study contributes to the discussion of how education through entrepreneurship is organised as an accelerator programme to help start-ups in a rural context.

Commercializing biomedical discoveries is a challenging process for many reasons. However, Academic Medical Centers (AMC) that have teaching, patient care, research, and service engrained in their mission are well poised to host these discoveries. These academic discoveries can lead to improvement in patient health and economic development if supported to cross the “valley of death” through institutional assistance, by providing guidance, gap funding and product development expertise. Colorado has a vibrant local startup ecosystem, state support for commercialization and entrepreneurship as well as critical mass of product development expertise. University of Colorado Anschutz Medical Campus, as a major AMC, is an engine for growth for the region. This article discusses innovation efforts at the University of Colorado Anschutz Medical Campus as a case study, which is built around two major efforts: the CCTSI and CU Innovations. I-Corps at CCTSI and the SPARK|REACH program of CU Innovations have been instrumental in fostering innovation, commercialization, and entrepreneurship on the campus.

Gesture recognition is critical in the field of Human-Computer Interaction, especially in healthcare, rehabilitation, sign language translation, etc. Conventionally, the gesture recognition data collected by the inertial measurement unit (IMU) sensors is relayed to the cloud or a remote device with higher computing power to train models. However, it is not convenient for remote follow-up treatment of movement rehabilitation training. In this paper, based on a field-programmable gate array (FPGA) accelerator and the Cortex-M0 IP core, we propose a wearable deep learning system that is capable of locally processing data on the end device. With a pre-stage processing module and serial-parallel hybrid method, the device is of low-power and low-latency at the micro control unit (MCU) level, however, it meets or exceeds the performance of single board computers (SBC). For example, its performance is more than twice as much of Cortex-A53 (which is usually used in Raspberry Pi). Moreover, a convolutional neural network (CNN) and a multilayer perceptron neural network (NN) is used in the recognition model to extract features and classify gestures, which helps achieve a high recognition accuracy at 97%. Finally, this paper offers a software-hardware co-design method that is worth referencing for the design of edge devices in other scenarios.

Background Boron neutron capture therapy (BNCT) utilizes tumor-selective particle radiation. This study aimed to assess the safety and efficacy of accelerator-based BNCT (AB-BNCT) using a cyclotron-based neutron generator (BNCT 30) and 10B-boronophenylalanine (SPM-011) in patients with recurrent malignant glioma (MG) (primarily glioblastoma [GB]). Methods This multi-institutional, open-label, phase II clinical trial involved 27 recurrent MG cases, including 24 GB cases, who were enrolled from February 2016 to June 2018. The study was conducted using the abovementioned AB-BNCT system, with 500 mg/kg SPM-011 (study code: JG002). The patients were bevacizumab-naïve and had recurrent MG after standard treatment. The primary endpoint was the 1-year survival rate, and the secondary endpoints were overall survival (OS) and progression-free survival (PFS). Results were compared to those of a previous Japanese domestic bevacizumab trial for recurrent GB (JO22506). Results The 1-year survival rate and median OS of the recurrent GB cases in this trial were 79.2% (95% CI: 57.0–90.8) and 18.9 months (95% CI: 12.9–not estimable), respectively, whereas those of JO22506 were 34.5% (90% CI: 20.0–49.0) and 10.5 months (95% CI: 8.2–12.4), respectively. The median PFS was 0.9 months (95% CI: 0.8–1.0) by the RANO criteria. The most prominent adverse event was brain edema. Twenty-one of 27 cases were treated with bevacizumab following progressive disease. Conclusions AB-BNCT demonstrated acceptable safety and prolonged survival for recurrent MG. AB-BNCT may increase the risk of brain edema due to re-irradiation for recurrent MG; however, this appears to be controlled well with bevacizumab.

Starta uzņēmējdarbības vadība Vadība, administrēšana un nekustamo īpašumu pārvaldība Startup Entrepreneurship Management, Administration and Real Estate Management Pasaule pēdējo 10-20 gadu laikā ir piedzīvojusi strauju jaunuzņēmumu skaita un ekosistēmas palielināšanos un no 2010. gada līdz šim brīdim jaunuzņēmumu skaits Eiropā ir palielinājies vairāk nekā 5 reizes. Jaunuzņēmumu videi piedzīvojot strauju izaugsmi Latvijai būtu svarīgi būt konkurētspējīgai šajā vidē, bet pašlaik tā dažādos radītājos ievērojami atpaliek no kaimiņvalstīm un daudzām citām Eiropas valstīm. Jaunuzņēmumiem ir liels potenciāls uz strauju izaugsmi un valsts ekonomikas uzlabošanu, un līdz ar to ir svarīga spēcīga vide. Redzot ievērojami mazākus radītājus dažādos jaunuzņēmumu datos, darba mērķis ir pamatojoties uz zinātniskās literatūras izpēti, datu ievākšanu un analīzi, sniegt secinājumus un ieteikumus par Latvijas jaunuzņēmumu vides pilnveidošanas iespējām. Darbā tiek izmantotas 4 pētnieciskās metodes – monogrāfiskā metode, kura tiks izmantota apskatot pieejamo literatūru un avotus un tos aprakstot, sekundāro datu analīzes metode, kura tiks izmantota, lai analizētu datus un trendus par jaunuzņēmumiem Latvijā, Igaunijā, Vācijā, salīdzinošā metode kura tiks izmatota, lai salīdzinātu jaunuzņēmumu vidi un atbalsta rīkus Latvijā, Igaunijā, Vācijā un kvalitatīva metode tiks izmantota, lai izveidotu jautājumus un pēc tam intervētu jaunuzņēmumu pārstāvjus Latvijā. Darba rezultāti parāda to, ka Latvijas jaunuzņēmumu vide pašlaik ir paaudzi iepriekš nekā Igaunija vai Vācija, kas nozīmē to, ka Latvijā nav izaudzis neviens liels jaunuzņēmums kurš tālāk investē jaunos projektos. Rezultāti parāda arī to, ka Latvijā ir pārāk augsta birokrātija, nav pietiekami daudz atbalsta rīki, galvenokārt kvalitatīvi akseleratori. Rezultāti arī parāda ka sadarbība starp jaunuzņēmumu nozari, valdību, universitātēm pašlaik nav pietiekamā līmenī, lai veicinātu pietiekamu vides attīstību un no valdības puses netiek finansēta neviena studiju vieta jaunuzņēmumu studiju programmās. Starp galvenajiem priekšlikumiem darbā Latvijas investīciju aģentūrai tika izvirzīts priekšlikums vienkāršot jaunuzņēmumu dokumentācijas procesu un apjomu priekš finansējuma saņemšanas un uzskaites. Tika izvirzīts arī priekšlikums Latvijas jaunuzņēmumu vides veidotājiem izveidot jauns, kvalitatīvus atbalsta rīkus, galvenokārt akseleratorus, kā arī Izglītības un zinātnes ministrijai pārvirzīt daļu no valsts finansētajām studiju vietām uz jaunuzņēmumu studiju programmām. Darbs sastāv no 68 lappusēm, 95 literatūras avotiem un 7 pielikumiem. The world has seen a rapid increase in the number of start-ups and a major growth in the ecosystem over the last 10-20 years, and the number of start-ups in Europe has grown more than five times since 2010. With the environment of start-ups experiencing rapid growth, it would be important for Latvia to be competitive in this environment, but currently it lags far behind its neighbors and many other European countries in various aspects. Start-ups have great potential for rapid growth and improvement of the national economy, and thus a strong environment is important. Seeing significantly smaller numbers in various data of start-ups, the aim of the work is to provide conclusions and recommendations on the possibilities of improving the environment of Latvian start-ups based on the research of scientific literature, data collection and analysis. 4 research methods are used in the work – monographic method, which will be used by reviewing available literature and sources and describing them, secondary data analysis method, which will be used to analyze data and trends on start-ups in Latvia, Estonia, Germany, comparative method, which will be used, to compare the start-up environment and support tools in Latvia, Estonia, Germany and a qualitative method will be used to create questions and then interview start-up representatives in Latvia. The results of the work show that the environment of Latvian start-ups is currently a generation behind Estonia or Germany, which means that no large start-ups have grown in Latvia that further invest in new projects. The results also show that there is too much bureaucracy in Latvia, not enough support tools, mainly high-quality accelerators. The results also show that cooperation between the start-up industry, the government, universities is currently not at a sufficient level to promote sufficient environmental development and no study places in start-up study programs are funded by the government. Among the main proposals in the work are as follows: Latvian Investment Agency a proposal to simplify the documentation process and volume for start-ups for receiving financing. A proposal was also made for Latvian start-up environment creators to create new, high-quality support tools, mainly accelerators, as well as for the Ministry of Education and Science to redirect part of the state- funded study places to start-up study programs. The work consists of 68 pages, 95 references and 7 appendices.

Pedestrian detection is one of the key problems in the emerging self-driving car industry. In addition, the Histogram of Gradients (HOG) algorithm proved to provide good accuracy for pedestrian detection. Many research works focused on accelerating HOG algorithm on FPGA (Field-Programmable Gate Array) due to its low-power and high-throughput characteristics. In this paper, we present an energy-efficient HOG-based implementation for pedestrian detection system on a low-cost FPGA system-on-chip platform. The hardware accelerator implements the HOG computation and the Support Vector Machine classifier, the rest of the algorithm is mapped to software in the embedded processor. The hardware runs at 50 Mhz (lower frequency than previous works), thus achieving the best pixels processed per clock and the lower power design.

Convolutional Neural Networks (CNNs) are a very popular class of artificial neural networks. Current CNN models provide remarkable performance and accuracy in image processing applications. However, their computational complexity and memory requirements are discouraging for embedded real-time applications. This paper proposes a highly optimized CNN accelerator for FPGA platforms. The accelerator is designed as a LeNet CNN architecture focusing on minimizing resource usage and power consumption. Moreover, the proposed accelerator shows more than 2x higher throughput in comparison with other FPGA LeNet accelerators with reaching up 14 K images/sec. The proposed accelerator is implemented on the Nexys DDR 4 board and the power consumption is less than 700 mW which is 3x lower than the current LeNet architectures. Therefore, the proposed solution offers higher energy efficiency without sacrificing the throughput of the CNN.

As big data analytics systems are squeezing out the last bits of performance of CPUs and GPUs, the next near-term and widely available alternative industry is considering for higher performance in the data center and cloud is the FPGA accelerator. We discuss several challenges a developer has to face when designing and integrating FPGA accelerators for big data analytics pipelines. On the software side, we observe complex run-time systems, hardware-unfriendly in-memory layouts of data sets, and (de)serialization overhead. On the hardware side, we observe a relative lack of platform-agnostic open-source tooling, a high design effort for data structure-specific interfaces, and a high design effort for infrastructure. The open source Fletcher framework addresses these challenges. It is built on top of Apache Arrow, which provides a common, hardware-friendly in-memory format to allow zero-copy communication of large tabular data, preventing (de)serialization overhead. Fletcher adds FPGA accelerators to the list of over eleven supported software languages. To deal with the hardware challenges, we present Arrow-specific components, providing easy-to-use, high-performance interfaces to accelerated kernels. The components are combined based on a generic architecture that is specialized according to the application through an extensive infrastructure generation framework that is presented in this article. All generated hardware is vendor-agnostic, and software drivers add a platform-agnostic layer, allowing users to create portable implementations.

International audience; Biomedical applications at high-energy particle accelerators have always been an important section of the applied nuclear physics research. Several new facilities are now under constructions or undergoing major upgrades. While the main goal of these facilities is often basic research in nuclear physics, they acknowledge the importance of including biomedical research programs and of interacting with other medical accelerator facilities providing patient treatments. To harmonize the programs, avoid duplications, and foster collaboration and synergism, the International Biophysics Collaboration is providing a platform to several accelerator centers with interest in biomedical research. In this paper, we summarize the programs of various facilities in the running, upgrade, or construction phase.

Recently, Deep Neural Networks (DNNs) have been widely used in natural language processing. However, DNNs are often computation-intensive and memory-expensive. Therefore, deploying DNNs in the real world is very difficult. In order to solve this problem, we proposed a network model based on the dilate gated convolutional neural network, which is very hardware-friendly. We further expanded the word representations and depth of the network to improve the performance of the model. We replaced the Sigmoid function to make it more friendly for hardware computation without loss, and we quantized the network weights and activations to compress the network size. We then proposed the first FPGA (Field Programmable Gate Array)-based event detection accelerator based on the proposed model. The accelerator significantly reduced the latency with the fully pipelined architecture. We implemented the accelerator on the Xilinx XCKU115 FPGA. The experimental results show that our model obtains the highest F1-score of 84.6% in the ACE 2005 corpus. Meanwhile, the accelerator achieved 95.2 giga operations (GOP)/s and 13.4 GOPS/W in performance and energy efficiency, which is 17/158 times higher than the Graphics Processing Unit (GPU).

Objective: Conflicts have negative impacts on organizational performance and can lead to company mortality. The GVentures Accelerator, from the School of Business Administration of São Paulo (EAESP-FGV) identified that several startups that failed during and after the acceleration process had conflicts among the founding partners. This work aims to understand why some entrepreneurial teams in a pre-seed stage are able to overcome conflicts while others are not Methodology: We conducted a case study of 9 accelerated startups using in-depth interviews with 20 founding partners, and the interviews were followed by the codification and analysis of the cases with support from the manager responsible for the accelerator.Results: The research concludes that operational conflicts that escalate to affective conflicts due to disagreements in the process of giving and receiving feedback and/or mistrust between partners can lead to the dissolution of a company. On the other hand, founding members, even if they experience affective conflicts, are able to overcome the problems using the strategies of taking a step aside, giving in and putting their egos aside. It was also identified that the acceleration process tends to exacerbate the operational conflicts between founding partners.Theoretical/methodological contributions: The research contributes to the literature about founding teams by pointing out that operational conflict does not necessarily lead to the dissolution of the organization or closing the business, but operational conflicts that intensify to affective conflicts due to disagreements in the process of giving and receiving feedback and/or distrust among partners can lead to the dissolution of the organizationRelevance/originality: Founding teams are the backbone of any company. In spite of several articles discussing team conflict little is known about why some entrepreneurial team are able to overcome conflicts while other not. Moreover, also little is known about the role accelerators play in these conflicts.Social/management contribution: The conclusions about the strategies for managing founding team conflicts: taking a step aside, giving in and putting their egos aside; may be very useful for both entrepreneurs and accelerators’ management team in dealing with conflicts among founders.

In the field of beam physics, two frontier topics have taken center stage due to their potential to enable new approaches to discovery in a wide swath of science. These areas are: advanced, high gradient acceleration techniques, and x-ray free electron lasers (XFELs). Further, there is intense interest in the marriage of these two fields, with the goal of producing a very compact XFEL. In this context, recent advances in high gradient radio-frequency cryogenic copper structure research have opened the door to the use of surface electric fields between 250 and 500 MV/m. Such an approach is foreseen to enable a new generation of photoinjectors with six-dimensional beam brightness beyond the current state-of-the-art by well over an order of magnitude. This advance is an essential ingredient enabling an ultra-compact XFEL (UC-XFEL). In addition, one may accelerate these bright beams to GeV scale in less than 10 meters. Such an injector, when combined with inverse free electron laser-based bunching techniques can produce multi-kA beams with unprecedented beam quality, quantified by ~50 nm-rad normalized emittances. These beams, when injected into innovative, short-period (1-10 mm) undulators uniquely enable UC-XFELs having footprints consistent with university-scale laboratories. We describe the architecture and predicted performance of this novel light source, which promises photon production per pulse of a few percent of existing XFEL sources. We review implementation issues including collective beam effects, compact x-ray optics systems, and other relevant technical challenges. To illustrate the potential of such a light source to fundamentally change the current paradigm of XFELs with their limited access, we examine possible applications in biology, chemistry, materials, atomic physics, industry, and medicine which may profit from this new model of performing XFEL science., 80 pages, 24 figures

The upgrade of the Large Hadron Collider (LHC) to the High Luminosity LHC (HL-LHC) is required to probe the physics beyond Standard Model. After the ongoing long shutdown (LS2), the accelerator luminosity will be increased 2 to 3 times as compared to designed luminosity value i.e. 10^34 cm-2s-1 . To meet the requirements of higher rates environment of HL-LHC era, the muon system of ATLAS detector needs to be upgraded. Therefore, the small wheel comprised of Cathode Strip Chambers (CSC), monitored Drift Tubes (MDT) chambers and Thin Gap Chambers (TGC) will be replaced by the New Small Wheel (NSW). The NSW will be constituted by MicroMegas gaseous detectors (from the MPGD family) and small-strip Thin Gap Chambers (sTGC). Micromegas detectors will be used for tracking as well as triggering purpose. For each of the two NSW, 16 modules will be installed in 16 sectors i.e. 8 large sectors and 8 small sectors; covering total area of ~ 1200 m2, each detector with an individual area between 2 and 3 m2. Micromegas are ionization-based gaseous detectors made up of parallel plates, having a thin amplification region separated from the conversion region via a thin metallic mesh. The production of Micromegas detectors as well as methods adopted to achieve required geometrical precision are presented. Specific measurement devices have been developed in the last few years to determine the mechanical metrology quality of Micromegas chambers, required for NSW. Planarity measurements of readout panels as well as module after assembly are done with a specifically developed co-ordinate measuring machine (CMM). The methodology to obtain such results as well as obtained results will be presented. Results of in-plane measurements (XY co-ordinate) performed using Rasnik masks etched on the PCBs are also reported. Modules constructed using the dedicated tools and methods are validated using cosmic muons, performances of the detectors will also be shown.

In this paper, we present an overview of the European Processor Initiative (EPI), one of the cornerstones of the EuroHPC Joint Undertaking, a new European Union strategic entity focused on pooling the Union’s and national resources on HPC to acquire, build and deploy the most powerful supercomputers in the world within Europe. EPI started its activities in December 2018. The first three years drew processor and platform designers, embedded software, middleware, applications and usage experts from 10 EU countries together to co-design Europe’s first HPC Systems on Chip and accelerators with its unique Common Platform (CP) technology. One of EPI’s core activities also takes place in the automotive sector, providing architectural solutions for a novel embedded high-performance computing (eHPC) platform and ensuring the overall economic viability of the initiative.

In today’s market era of rapid technological innovation even the most established agribusiness players risk being left behind in the wake of disruptive technology. How can large agriculture companies keep up? One emerging possibility for corporate innovation is that of the corporate start-up accelerator. Companies across all sectors, including agricultural and food technology, are increasingly using these programs to tap into early innovation. In 2017, Alltech ran its first program, the Pearse Lyons Accelerator for agri-business and ag-tech start-ups. This article serves to describe and analyze how corporate accelerators have populated start-up ecosystems and how the ag-tech and food sectors are reflecting this trend. Recommendations based on experience and critical feedback are presented to those considering developing their own accelerators and the role it might play in responsible innovation.

Introduction. Modern trends in the development of education and society as a whole not only strengthen the role of modern universities in transformation of the social sphere, but also require a new approach to developing skills and over-professional competencies of university students, especially students of social and humanitarian specialties, who will live in conditions of a new technological paradigm. In this connection, classical universities have a special role because social and cultural development of society is their original mission. Therefore, development of social and entrepreneurial competence of students is an urgent educational task of modern classical universities. Education of social entrepreneurs is one of the relevant educational tasks which solution enables to provide additional resources for growth of welfare of fellow citizens. The work of social entrepreneurs is directed to the creative solution of social problems of the region, the help to its residents and positive changes in local society. The aim of the article is to demonstrate the experience of approbation of the educational technology for developing social and entrepreneurial competence of students in a classical university. Methodology and research methods. The methodological basis of the work is based on system-activity and competence approaches. The authors have used such methods as analysis of scientific literature; analysis of documents; desktop research; interviewing and questioning of participants in the educational process; educational experiment; reflection and generalisation of practical experience of implementing a program of accelerating business projects in a classical university (case study). Results and scientific novelty. The new approach to the development of social and entrepreneurial competence of students is proposed by the authors; that approach is based on the introduction of “Social Accelerator” technology into educational practice of higher education institutions. The technology structure is described; as obligatory components, the structure includes: short-term (several months long) program of training of the beginning entrepreneurs; mentor maintenance of students’ start-ups promotion; creation of conditions to present students’ project products to the public. The summary of the “Social Accelerator” functioning cycle is given. The results of its approbation carried out from 2015 to 2017 on the platform of the Park of Social and Humanitarian Technologies of the Tomsk State University are presented. The analysis of students’ responses proved emergence of a sustained interest to social and reformative activity and desire to continue training in the chosen direction at city, regional, all-Russian and international schools and programs. Practical significance . The “Social Accelerator” technology proposed by the authors is the innovative effective instrument of generation of a new wave of entrepreneurs; the technology can be used for students’ education at classical universities. The technology allows to: timely distinguish the persons who are capable of business activity; make active their initiative; provide timely assistance in development of enterprise tendencies and obtaining necessary knowledge.

Convolutional neural networks (CNNs) are widely used in modern applications for their versatility and high classification accuracy. Field-programmable gate arrays (FPGAs) are considered to be suitable platforms for CNNs based on their high performance, rapid development, and reconfigurability. Although many studies have proposed methods for implementing high-performance CNN accelerators on FPGAs using optimized data types and algorithm transformations, accelerators can be optimized further by investigating more efficient uses of FPGA resources. In this paper, we propose an FPGA-based CNN accelerator using multiple approximate accumulation units based on a fixed-point data type. We implemented the LeNet-5 CNN architecture, which performs classification of handwritten digits using the MNIST handwritten digit dataset. The proposed accelerator was implemented, using a high-level synthesis tool on a Xilinx FPGA. The proposed accelerator applies an optimized fixed-point data type and loop parallelization to improve performance. Approximate operation units are implemented using FPGA logic resources instead of high-precision digital signal processing (DSP) blocks, which are inefficient for low-precision data. Our accelerator model achieves 66% less memory usage and approximately 50% reduced network latency, compared to a floating point design and its resource utilization is optimized to use 78% fewer DSP blocks, compared to general fixed-point designs.

With the growing demand from scientific projects such as the X-ray free electron laser (XFEL), ultrafast electron diffraction/microscopy (UED/UEM) and electron ion collider (EIC), the semiconductor photocathode, which is a key technique for a high brightness electron source, has been widely studied in China. Several fabrication systems have been designed and constructed in different institutes and the vacuum of most systems is in the low 10−8 Pa level to grow a high QE and long lifetime photocathode. The QE, dark lifetime/bunch lifetime, spectral response and QE map of photocathodes with different kinds of materials, such as bialkali (K2CsSb, K2NaSb, etc.), Cs2Te and GaAs, have been investigated. These photocathodes will be used to deliver electron beams in a high voltage DC gun, a normal conducting RF gun, and an SRF gun. The emission physics of the semiconductor photocathode and intrinsic emittance reduction are also studied.

We present a scientific computing accelerator on FPGA that uses hundreds of processors working in parallel organized in several SIMD cores. The accelerator is installed within an Ethernet network and acts as a high-performance computing server. A prototype is presented for processing solar images and achieves a great performance that can compete with a cluster. © 2019 Copyright is held by the owner/author(s). Publication rights licensed to ACM., This work has been partially funded by the Spanish Ministerio de Ciencia, Innovación y Universidades, through Projects No.ESP2016-77548-C5-1-R and RTI2018-096886-B-C51, including a percentage from European FEDER funds. Authors also acknowledge financial support from the State Agency for Research of the Spanish MCIU through the >Center of Excellence Severo Ochoa> award to the Instituto de Astrofísica de Andalucía (SEV-2017-0709)

Start-ups are unique in that they are young, innovative, and growth-oriented entrepreneurial ventures. They have high potential benefits, but come with great uncertainty. Uncertainty arises from the complexity of strategic choices, the limited abilities of the entrepreneurs, and a lack of legitimacy. However, if these challenges are overcome through an iterative learning process, start-ups provide value to shareholders and society by introducing meaningful innovations, such as solutions to climate change. To defy the slim odds of success, business incubation has been heralded to increase the performance of start-ups. Incubation offers a package of services to start-ups, with the general aim to improve their development. It does so in an assertive manner, actively interfering with the start-up’s development. However, whether and why incubation structurally contributes to the performance of start-ups has been a topic of debate among both academics as well as practitioners. Therefore, to discern the noble goals and high hopes from what really is going on, this thesis addresses the following main research question: How does business incubation influence the performance of start-ups? The answer to this question emerged from four studies. The first systematically reviewed the network-based incubation literature. It showed that incubation clearly provides a range of intermediary benefits, such as valuable business knowledge and a sense of belonging. However, whether and under which conditions these intermediary benefits lead to increased start-up performance remains ambiguous. Thereby, study 1 laid out the background of literature to this thesis and reiterated the need for more empirical research. Study 2 tested the influence of incubation on an original dataset of 269 start-ups of which some were, and some were not accepted to two Dutch incubators. Controlling for pre-incubation quality to deal with the non-random selection, the study shows that incubation increases the performance of start-ups. While this finding provided an important element to the answer of the main research question, a convincing explanatory mechanism was still lacking. The third study therefore took a longitudinal approach. Comparing the experimentation processes of start-ups that were influenced by incubation with start-ups that were not, this study showed that depending on the phase of the start-up’s experimentation process and associated cognitive biases of the start-up founders, the incubator provides particular types of services. Without having to know the end solution, incubation programmes can improve the start-up’s experimentation process, and thus its performance. Study 4 took a quantitative approach to study the digital networks of incubated start-ups. It showed that incubated start-up benefit from a large share of relations to people with very different knowledge and skills than the founders of start-up. Similarly, they benefit from a large share of relations to people from a similar institutional background. Incubators thus can to take care of the start-up’s networks. Coming back to the research question, this thesis shows that incubation is about relieving the start-up from some cognitive biases through a tailored and dynamic mix of services. Thereby, incubation leads to significant and positive impact on start-up performance.