My research objective is to advance the state-of-the-art in time series forecasting by exploring the design of novel deep neural network models that take advantage of the distinctive properties of sequential data. I aim to identify and incorporate key features of time series, such as temporal dependencies and trends, into the architecture of my models to enhance their accuracy and generalizability. By combining my domain knowledge in time series analysis and machine learning with cutting-edge deep learning techniques, I aim to develop more powerful and effective forecasting models that can be applied to a wide range of real-world problems.
Koohfar, Sepideh, and Laura Dietz. 2022. "Adaptive Temporal Attention Mechanism to Address Distribution Shifts". NeurIPS 2022 Workshop on Robustness in Sequence Modeling. https://openreview.net/pdf?id=5aIwxkn0MzC. Appendix.
Koohfar, Sepideh and Laura Dietz. 2022. "Adjustable Context-aware Transformer". 7th Workshop on Advanced Analytics and Learning on Temporal Data, AALTD@ECML. https://project.inria.fr/aaltd22/files/2022/08/AALTD22_paper_4854.pdf. Appendix.
Computer Science PhD, University of New Hampshire, NH, Expetced to graduate by December 2023
Bachelor of Information technology, University of Isfahan, Iran