Skip to content

Performed literature survey on various architectures like FFNN, RNN, LSTM RNN, Gated RNN, and Transformers (SOTA Model) - unidirectional and bidirectional versions of all these, that can be used to solve univariate time series problems.

Notifications You must be signed in to change notification settings

khetansarvesh/Time-Series-Modelling

Repository files navigation

Time-Series-Modelling

To understand theory go to my notes at following link (recommended) : https://drive.google.com/drive/folders/1y3VNtNawi8j7H_42IPRS2xVeNl_mRqYm?usp=sharing

In my notes I have given a detailed explaination of how one can solve the univariate time series problem using even a FFNN but since there exists a problem in solving with FFNN it motivated the idea of RNN, what is this problem?? Read my notes to understand same. Now there exists a problem with RNN i.e. vanishing and exploding gradients which motivates the idea of LSTMs and since these exists a problem with LSTMs it motivated the idea of GRUs. To solve the parallel computation problem with GRUs the idea of transformer based models poped up.

LSTM

4 hidden layer stacked LSTM RNN architecture to solve the univariate time series forcasting problem of google stock price prediction using pytorch deep learning library.

Transformer-Encoder

About

Performed literature survey on various architectures like FFNN, RNN, LSTM RNN, Gated RNN, and Transformers (SOTA Model) - unidirectional and bidirectional versions of all these, that can be used to solve univariate time series problems.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published