Adapting R3GAN (NeurIPS 2024) to multivariate time series imputation: thesis model, GAN/Transformer baselines, an interactive Dash GUI, and an empirical study (FMGAN).

This is a warehouse for TimeSeries-pytorch-models, can be used to train your dataset

Deep learning models for classifying powder materials using multi-wavelength optical scattering signals. Achieves 99% accuracy with TimesNet on RGB light scattering time series.

FFT-guided period selection + 2D Inception CNN + embedding-aware adaptive probabilistic head for multivariate time-series forecasting.

💸계절성 있는 시계열 데이터 기반 기업 미래 매출 예측하기

Research-only DL sequence modeling (ToyTimesNet-style) for XAUUSD time series. Windowing, training, and evaluation under academic constraints.