A new research paper explores various models for forecasting time series data from the Lomnicky Stit neutron monitor. The study compares simple seasonal baselines with advanced deep learning architectures like LSTM, TCN, N-BEATS, and Kolmogorov-Arnold Networks (KAN), alongside two quantum-inspired variants, QiLSTM and QiKAN. Results indicate that the quantum-inspired QiKAN model achieved the lowest forecasting error, though the basic Seasonal Naive model also performed competitively. The findings suggest that models with strong seasonal or low-dimensional functional priors are effective for periodic scientific signals. AI
IMPACT Suggests parsimonious models with strong priors can match or outperform complex architectures for periodic scientific signals.
RANK_REASON The cluster contains an academic paper detailing novel research findings in time series forecasting. [lever_c_demoted from research: ic=1 ai=1.0]
- Kolmogorov-Arnold Networks
- long short-term memory
- N Beats Neural Basis Expansion Analysis For Time Series Forecasting
- QiKAN
- QiLSTM
- Temporal Convolutional Network
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
