This directory contains Python script to run linear dynamical system clustering experiments locally without dependencies on Google packages.
The files in this directory are meant to be run as python scripts. To install
dependencies, run pip install -r requirements.txt If there are error messages
related to the pylds package, do the following: 1. Install numpy through pip install numpy 2. Install Cython through pip install Cython 3. Remove pylds
from the requirements file, and install the rest requirements pip install -r requirements.txt 4. Install pylds through pip install pylds
Example to run experiments for learning eigenvalues: python
experiment_learn_eig.py --output_dir=may11_eig_2d_test/
--min_seq_len=20 --max_seq_len=1000 --num_sampled_seq_len=2
--num_repeat=1 --hidden_dim=2 --true_eig=0.1,0.8
Example to run experiments for running clustering experiments: python
experiments.py --output_dir=may19_3d --hidden_state_dim=3
--min_seq_len=100 --max_seq_len=2000 --num_sampled_seq_len=20
--num_systems=100 --num_repeat=100
--cluster_center_dist_lower_bound=0.1 --hide_inputs=true
For more information about the flags, please see documentation in the scripts.