A business / market opportunity assessment for hardware concepts & device designs for deep learning (DL)

我们正在开发计算硬件概念和设计，与现有的深度学习（DL）商品硬件相比，这些硬件可以提供两到三个数量级的数据处理能力和更好的能源效率。我们的技术使DL复杂性的增加，使其实际部署DL的方式目前还不可能。我们的目标是可扩展的硬件，可以针对各种设备和应用程序进行定制。其好处是：1）在桌面/数据中心，我们将通过实现更大和更复杂的DL网络的部署以及以更快的速度处理更多输入数据来实现DL的进一步创新和复杂化，2）对于移动的或嵌入式应用，我们将允许i）部署更复杂和准确的DL网络，ii）能耗的急剧减少并因此改善了移动的设备的正常运行时间，以及iii）功率需求的减少使得系统在各种情况下更加便携和可用（例如，自动驾驶车辆、个人辅助等）。我们的设计是新颖和独特的，并已被证明优于同类最好的硬件设计提出了迄今为止和商业上可用的解决方案2至3个数量级。我们的设计可以为高端台式机/服务器市场或嵌入式/移动的/物联网领域提供下一代DL解决方案。这是一份I2 I市场评估资助申请，旨在深入现场和潜在客户，以便适当地确定适当的应用程序，从而确定应作为我们创新技术成功商业化的第一步的配置。

We are developing computing hardware concepts and designs that deliver two to three orders of magnitude faster data processing capability and better energy efficiency than existing commodity hardware for Deep Learning (DL). Our techniques enable an increase in DL sophistication making it practical to deploy DL in ways not currently possible. We target scalable hardware that can be tailored to various devices and applications. The benefits are: 1) at the desktop/data-center, we will enable further innovation and sophistication in DL by enabling the deployment of larger and more sophisticated DL networks and the processing of more input data and at faster speeds, 2) for mobile or embedded applications, we will allow i) the deployment of more sophisticated and accurate DL networks, ii) a drastic reduction in energy consumption and thus improvement in up-time for mobile devices, and iii) a reduction in power needs making the system more portable and usable in a variety of scenarios (e.g., autonomous vehicles, personal assists, etc.).Our designs are novel and unique and have been shown to outperform the best in class hardware designs proposed thus far and commercially available solutions by 2 to 3 orders of magnitude. Our designs canenable the next generation of DL solutions be it for the high-end desktop/server market or the embedded/mobile/IoT space. This is an application for an I2I Market Assessment grant to reach out to the field and potential customers in order to appropriately identify appropriate applications and thus the configurations that should be targeted as a first step toward successful commercialization of our innovative technology.