Skip to content

Latest commit

 

History

History
20 lines (13 loc) · 705 Bytes

File metadata and controls

20 lines (13 loc) · 705 Bytes

Time2Vec: Learning a Vector Representation of Time

This is a try to implement of the paper Time2Vec: Learning a Vector Representation of Time in TensorFlow 2 & Keras.

Prerequisites

At least TensorFlow (Tested on Python 3.6.9 & TF 2.2.0) for using layers.py or models.py directly.

For run.py, numpy 1.18.5 & pandas 1.1.1 needed.

Data used

Currently tested on dataset: Water Levels in Venezia, Italia on Kaggle. Download from Kaggle and extract all files to Time2Vec/data.

TODO:
  1. Test on more dataset including those used in the paper to ensure the model correctness.