mhoward3210 / ccl

GNU General Public License v2.0
3 stars 4 forks source link

Add second file for comparison of learning constraint implementations #5

Closed joaomoura24 closed 3 years ago

joaomoura24 commented 6 years ago

In a previous pull request I added a file demo_N_estimate_comp.m where I used our data (kuka wiping experiment) and attempt to estimate the constraint using the methods of our 2 code bases - the idea is to solve the same problem with both implementations with the goal of gradually converge the 2 code bases.

We got a bug when running this file that is still unsolved. To help solving this I added now a second file: demo_N_estimate_comp2.m that uses the data from the 2D toy example (KCL code base) and estimates the Projection matrix N using KCL method (implemented in function learn_alpha_ccl.m) and our method (implemented in function def_constraint_estimator.m). For this case the code runs and gives the right result.

In order to run this file, we have first to generate the data set. So, I also added a file in the data_generation folder called generate_data_toy_exampe_2D.m that generates the data for the toy example.

I am trying to keep separate the generation of data and the estimation of the constraint. So in principle 1 single file that estimates the constraint can run for any generated data (of any dimensionality) independently of the data.