I'm Fateme Roshani, but I preffer to be called Aida. I am a M.Sc. student in Artificial Intelligence at the University of Tehran and a graduate researcher at the Computational Audio-Vision Lab (CAVLab), supervised by Prof. Reshad Hosseini.
I'm interested in reliable machine learning for time-series data, especially learning with noisy labels and weak supervision. My broader interests include probabilistic forecasting, deep generative models, and reinforcement learning.
- Learning with noisy labels: Investigating robust time-series classification under imperfect annotations, including human-in-the-loop weak supervision and encoder-based label-refinement approaches.
- Time-series forecasting: Collaborating on a research toolkit for exploring encoder–generator architectures and evaluating forecasting models.
-
TimeGrad — Probabilistic Time-Series Forecasting
Co-developed a PyTorch reproduction of an autoregressive diffusion model for multivariate forecasting, with evaluations across five benchmark datasets. Collaborative project. -
Competitive Dynamic Pricing with Soft Actor-Critic
Co-developed a reinforcement learning agent for continuous pricing in a stochastic environment with changing demand and reactive competitors. Collaborative project. -
TurtleBot3 Global Localization with Particle Filters
Implemented and compared basic, adaptive, and MCMC particle filters for robot localization using ROS 2 and Gazebo. -
Multilingual Speech Language Identification
Co-developed a speech classification pipeline combining acoustic feature engineering, supervised learning, and unsupervised clustering.
I'm interested in research collaborations and international AI research internships.
- Email: fateme.roshani@ut.ac.ir
- LinkedIn: aida-roshani