lstm_seq2seq_model_prediction

Use the deep learning recursive neural network keras RNN-LSTM Seq2Seq Many to Many model to predict some untrained points on a circle.

Environment

Google Colab CPU/GPU/TPU
keras 2.2.5 LSTM
Ubuntu 18.04.3 LTS
Python 3.6.8
Numpy 1.17.3
Pandas 0.25.2
sklearn 0.21.3

Model

Result

By increasing the number of epochs and neurons, a satisfactory result is obtained in accuracy as shown above. This shows how important the adjustment of hyperparameters, including the selection of cost functions and activation functions, is in terms of accuracy.

Language

Explanation in Japanese