* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download 1 Title: Machine learning/Artificial intelligence for Prediction in
Survey
Document related concepts
Artificial intelligence for video surveillance wikipedia , lookup
Michael Tomasello wikipedia , lookup
Stephen Grossberg wikipedia , lookup
Machine learning wikipedia , lookup
Expert system wikipedia , lookup
Theoretical computer science wikipedia , lookup
Situated cognition wikipedia , lookup
Cognitive model wikipedia , lookup
Intelligence explosion wikipedia , lookup
Knowledge representation and reasoning wikipedia , lookup
Embodied cognitive science wikipedia , lookup
Philosophy of artificial intelligence wikipedia , lookup
Natural computing wikipedia , lookup
William Clancey wikipedia , lookup
Transcript
1 Title: Machine learning/Artificial intelligence for Prediction in Robotics (MAPiR) Robotics and Intelligent Systems group (ROBIN), University of Oslo Contact: Jim Tørresen, [email protected] and Kyrre Glette, [email protected] More information: http://www.mn.uio.no/ifi/english/research/groups/robin The PhD project will be connected to our ongoing Research Council of Norway funded projects about Robot Companions for Elderly Care and/or Prediction and Coordination for Robots and Interactive Music. They are both concerned with applying machine learning/ artificial intelligence for analyzing multi-sensor systems to train behavior models of both robots and human users for predicting future events and actions. Requirements: Applicants should have strong programming skills, and knowledge and experience with machine learning/artificial intelligence. 2 Main objective and summary of the project Humans are superior to computers and robots when it comes to being alert by applying multimodal sensing together with learned knowledge in predicting future events and choosing the best actions. Are we able to transfer these skills into intelligent systems and also apply that knowledge in how the systems interact with human users? The goal of this project is to design, implement and evaluate systems that can learn to predict and act using knowledge about human prediction and decision making mechanisms [1]. Computational models providing prediction are the target of this project. However, that requires an internal model of the system to be involved in the prediction whether it is a network of sensors, a data set, a robot, a human or a combination of two or more of these. Thus, we shall work on developing various kinds of internal models that can represent the behavior to be modeled (which together with perception provide a system with selfawareness capabilities) [5]. These would be relevant for robot control and user interaction. The applications would be within robotics (e.g. robot companions to be applied within elderly care or in other application areas) or to enhance the user adaptation of smartphones. The work would be building on the state-of-the-art knowledge in sensors, behavioral models, machine learning, and artificial intelligence. 3 Project background and scientific basis There is currently taking place a transition from systems earlier having a static behavior into now being able sense its user and being able to adapt to the current needs and preferences of its given user. This is relevant for many kinds of systems including robots which in the future are going to appear closer to humans than what we are used to; at home, at work and out in the society. Our research group is currently involved in several projects funded by the Research Council of Norway where we target to develop technology with such adaptive features. Thus, this PhD project is to complement with these and would be collaborating with researchers in them. The projects include Elderly Care with Robot Companion (MECS, 2016-2019) and Engineering Predictability with Embodied Cognition (EPEC, 2016–2019). The projects target to develop computational intelligence models being able to represent behaviour and act with timely response through prediction capabilities whether it regards the systems itself or its user. The foundation will be our earlier work undertaken on self-aware and self-expressive computing systems [3] including with motion studies [4]. 4 Research questions and scientific challenges We would like to focus on the following research challenges in the project: • Extract knowledge mechanisms of human prediction, intuition and reasoning • Develop learning methods for predicting future events and actions • Develop mechanisms for adaptive response from quick and instinctive to slower wellreasoned • Demonstrate the feasibility of the developed technology in a set of different use cases • Contribute to the development of a computational reference architecture 5 Scientific method The project will consist of working with relevant data collected from the selected applications and later in the project also testing the developed methods in real-time operation. Different available feature extraction and classification algorithms are to be compared, combined and extended including recent ones in artificial neural networks [2]. We will employ quantitative measures of performance, and supplement with statistical assessments, given the stochastic processes and real-world nature of the research. 6 Ethics There are few ethical issues related to the project but if present, approvals from the Regional Committees for Medical and Health Research Ethics would be collected, informed consent from all study participants, guaranteeing confidentiality and anonymous processing of data and acting in accordance with the Guidelines for Research Ethics. 7 Project timeline Year 1: Literature study. Choose relevant applications. Mandatory courses. Implement methods and first experiments based on simulation. Publication. Year 2-3: Results from real world application. More experiments with new or modified methods. Several publications. Year 4: Final experiments. An extensive publication. Thesis writing. • • • References [1] Evans, J. St. B. T. and Stanovich, K. E. Dual-process theories of higher cognition. Science, 8:223– 241, 2013. [2] LeCun, Y., Bengio, Y., Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444. [3] Lewis, P.R.; Chandra, A.; Funmilade, F.; Glette, K.; Chen, T.; Bahsoon, R.; Torresen, J.; Yao. X.: Arch. Aspects of Self-Aware and Self-Expressive Computing Syst.: From Psychology to Engineering, IEEE Computer, vol.48, no. 8, pp. 62-70, Aug. 2015 [4] Nymoen, K.,...Torresen J. Analyzing correspondence between sound objects and body motion. In ACM Trans. on Applied Perception, 10(2), 2013 [5] Schillaci, G., Hafner, V. V., & Lara, B. Exploration behaviours, body representations and simulations processes for the development of cognition in artificial agents. Frontiers in Robotics and AI, 3(June), 39, 2016.