(apologies in case of multiple receptions) The Luxembourg Institute of Science and Technology (LIST) is a Research and Technology Organization (RTO) active in the fields of materials, environment and IT. By transforming scientific knowledge into technologies, smart data and tools, LIST empowers citizens in their choices, public authorities in their decisions and businesses in their strategies.
https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.list.lu%2F&data=05%7C02%7Cuai%40engr.orst.edu%7C027b7599959242a2604d08dbf7d71aea%7Cce6d05e13c5e4d6287a84c4a2713c113%7C0%7C0%7C638376275830068280%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=3ePIq5acFDUoF4Ee3GsyxJh1fvN0WyQxBlk%2FqJMKj2w%3D&reserved=0 Time series analysis is vital for making informed decisions and improving the efficiency of systems, with significant financial implications. Applications range from logistics to healthcare, as well as understanding climate change. The context of this internship focuses more specifically on spatially distributed time series, such as data from weather or smart building sensors. Environmental sensor data often exhibits heavy spatial correlation. This means that the effective sample size will be much smaller than apparent. Therefore, using this data with sequence-to-sequence deep learning forecasting models is prone to distributional shift. A first step to address this issue is the development of tools to effectively characterize such distributional shift issues. Internship Contributions: * Characterize challenges related to distributional shift discrepancy metrics, non-stationarity of time series, and small effective sample sizes. * Survey and adapt techniques for time series visualization to highlight discrepancies in input and target time series. * Build proof-of-concept software to illustrate proposed adaptations, leveraging the Python gluonts library. Key Responsibilities: * Conduct a comprehensive survey of distributional shift discrepancy metrics and time series visualization primitives. * Identify and adapt suitable techniques for characterizing distributional shifts in time series data. * Develop and integrate software solutions into an existing codebase for environmental time series forecasting. * Collaborate with the research team to refine methods and contribute to ongoing projects. Education * Enrolled in Master or equivalent in Data Science, Statistics, Computer Science, or a related field. Experience and skills * Foundation in statistical modelling, machine learning, and data analysis * Proficiency in Python * Proficiency with Javascript/TypeScript is a plus * Familiarity with deep learning frameworks, preferably MXNet and/or PyTorch * Ideally, prior experience with time series analysis * Ability to work independently and in a team environment. Language skills * Good level both written and spoken English C1 More information at https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fapp.skeeled.com%2Foffer%2Fc%2F656e0d25de871be5b780fe5f%3Flanguage%3Den%26show_description%3Dtrue&data=05%7C02%7Cuai%40engr.orst.edu%7C027b7599959242a2604d08dbf7d71aea%7Cce6d05e13c5e4d6287a84c4a2713c113%7C0%7C0%7C638376275830068280%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=BxfuXmXSxVECEcCCjf0AN2lE%2FrZkj%2BtPnmo09qzfd9o%3D&reserved=0 Please apply using the form provided there. Best regards, Pierrick Bruneau Senior Research Associate
_______________________________________________ uai mailing list uai@engr.orst.edu https://it.engineering.oregonstate.edu/mailman/listinfo/uai
