epapagia/pointer-networks

An implementation of "Pointer Networks" in Tensorflow

Pointer Networks in Tensorflow

This is an implementation of Pointer Networks to solve the Convex Hull problem. Stacking RNN layers is supported.

Environments

• Python 3.x
• TensorFlow 1.2.x

Data

Convex Hull datasets such as "convex hull 5" and "convex hull 5-50" can be downloaded at Link.

Usage

training

$ python convex_hull.py --ARG=VALUE

evaluating

$ python convex_hull.py --forward_only=True --beam_width=VALUE --ARG=VALUE

visualizing

$ tensorboard --logdir=DIR

Results

Training on convex hull 5

Training on convex hull 5-50