EAGER: IIS: Empowering Probabilistic Reasoning with Random Projections

自动驾驶汽车等自主智能体需要在日益复杂和不确定的现实世界环境中智能和自适应地行动。为了科普真实的世界领域的不确定性和模糊性，人工智能系统严重依赖于统计方法。为了在不确定性下做出明智的决策，智能体需要对他们的环境进行概率推理。然而，在最坏的情况下，概率推理在计算上是非常困难的。虽然在过去几十年中取得了重大进展，但许多复杂的问题仍然无法解决。这个项目的目的是开发一个新的家庭的算法推理不确定性。这些新的技术有可能提供更有效的算法，决策，学习和推理，提高理论保证的准确性。这些技术将应用于广泛的领域，包括医疗诊断，信息提取，计算机视觉和机器人。本研究项目将开发一个新的家庭的算法，在不确定性的基础上随机投影推理。随机投影在数据挖掘和数据库系统的扩展中发挥了关键作用。虽然大大降低了计算成本，他们也提供了原则性的近似。本研究将探讨在概率推理的背景下使用基于通用散列方案的随机预测。该项目将开发在不确定性问题下进行学习和决策的新技术。具体而言，将开发具有改进的理论保证和实际性能的新框架和算法。为了提供有效的推理算法，随机投影的使用将被认为是与一系列现有的技术，包括现代优化，变分和抽样方法相结合。一个关键的重点将是开发实用技术和扩大到现实世界的领域。 所开发的技术将通过开放源码软件提供给学术界和工业界。教育和推广工作将包括让本科生参与独立的研究项目。

Autonomous agents such as self-driving cars are required to act intelligently and adaptively in increasingly complex and uncertain real-world environments. To cope with the uncertainty and ambiguity of real world domains, AI systems rely heavily on statistical approaches. To make sensible decisions under uncertainty, agents need to reason probabilistically about their environments. Probabilistic reasoning, however, is known to be computationally very difficult in the worst case. While significant progress has been made over the past decades, many complex problems remain out of reach. This project aims to develop a new family of algorithms for reasoning under uncertainty. These novel techniques have the potential to provide more efficient algorithms for decision-making, learning and inference with improved theoretical guarantees on the accuracy. These techniques will be applicable in a wide range of domains, including medical diagnosis, information extraction, computer vision, and robotics.This research project will develop a new family of algorithms for reasoning under uncertainty based on random projections. Random projections have played a key role in scaling up data mining and database systems. While drastically reducing computational cost, they also provide principled approximations. This research will explore the use of random projections based on universal hashing schemes in the context of probabilistic reasoning. The project will develop new techniques for learning and decision making under uncertainty problems. Specifically, new frameworks and algorithms with improved theoretical guarantees and practical performance will be developed. In order to provide efficient reasoning algorithms, the use of random projections will be considered in combination with a range of existing techniques, including modern optimization, variational, and sampling methods. A key focus will be to develop practical techniques and scale-up to real-world domains. The techniques developed will be made available to both academia and industry through open-source software. Educational and outreach efforts will include the involvement of undergraduate students undertaking independent research projects.