This repository pertains to our investigatory project that delves into the potential applications of implementing a predictive algorithm to forecast the weather, chances of rain in particular, using a quantum approach.
A quantum approaches need not always be faster but quantum algorithms are shown to exhibit significantly better results in comparison to classical approaches regarding problems whose answers are hard to calculate but easy to verify. Evidently, forecasting the chances of rain is one such problem.
The project aims to have a go at this problem by implementing an LSTM training model using quantum algorithms by implementing the LSTM gates using VQCs (Variational Quantum Circuits).
This could potentially see significant improvements in prediction and also could allow much larger datasets to train the model.
QLSTM capitalizes on quantum entanglement to reduce the number of parameters in the quantum circuit. This greatly reduces the noise and generates high expressive power.
~Here, xpressive power refers to the ability to represent certain functions or distributions with a limited number of parameters.
QLSTM learns faster and takes less epochs.
Stable convergence achievable in comparison to classical LSTM, particularly, spikes in loss function observed in classical LSTM is not observed in QLSTM.
Sriram B. Swami - 22BCE5103
Manasa Ganesh - 22BAI1262
Meghna Varma - 22BCE5085
AO Navin Kumar - 22BCE1020