Implemented 1D and multiple trajectories cases and wrote corresponding tests.

Raise error if the inputs for pde cases contain tuple, and write corresponding tests. Modified conftest.py: return additional t in data_3d_random_pde and data_5d_random_pde

Log-normal contour plot of RMSE between chaotic level and noise level for X and X dot.

…into lanyue_work

…into lanyue_work

… now.

…and avoids integrating the models for speed. Now is quite fast, and the plan is to use this to look at syntactical complexity.

… with syntax so far, which is surprising, but need to make sure it is plotting properly.

…ngs. Made a new script that loops over all the polynomial systems in the dysts database, and fits them (without noise). Also wrote a script to load in all the ODE functions as strings, and parse them to extract all the model coefficients. Needs some tuning, but so far looks like it is performing well compared to the true coefficients.

…ata was being scored without the corresponding ttest time range, so errors were not being computed correctly for the algorithm.

…score. Going to keep debugging this, and merge the main branch in case there are some bug fixes there.

…score. Going to keep debugging this, and merge the main branch in case there are some bug fixes there.

…parameters for each of the dynamical systems. Going to see how well it generalizes when I add in some error.

… analysis.

…into lanyue_work

…rge set of test trajectories near the attracting set.

…tion. Currently ode simulation taking too long with these long time series. Will need to think about this.

…s quite well even with 1 percent noise. Will polish this up for paper in a bit.

…erably off the attractor once the models are trained.

… with concluding about the dynamical properties.

…ge in main branch now to have access to miosr optimizer.

…eparate notebooks for now because computing everything is slow for some of the optimizers.

… branches and work on something else briefly.

…e main branch. Also moved the dysts-specific functionality (like computing the dynamical properties) to a pull request for the dysts library, which has now been merged. Removed a lot of the older files - if we need them later can grab from earlier commits. Still rerunning main Pareto scan script to clean it up and make sure it is working right.

…ost recent changes.

…into lanyue_work

… script now, STLSQ results are quite different than before. Chasing it down now.

…complexity.

…b to github so need to update the pre-commit file.

…tead of version 10.

Add words description to the notebook and removed unimportant plots as trials.

…d added strong_rmse option

…yue_work

…ning all the results at once so dont need multiple notebooks.

…for all the optimizers except MIOSR. Need to update the plotting notebook to use the new data. Made some small changes to the run_all script so that all of the optimizers can run weak form.

…ents were being reshuffled even when weak form was false.

…tebook.

…into lanyue_work

…umbers. The issue is that the true_coefficients matrix was reordered for the weak library, but the error matries were also reordered, so then they didnt match again. Now the true_coefficients are left alone. Reran the STLSQ results which are looking much more reasonable now.

…into lanyue_work

…into lanyue_work

… linting and pytest errors, this is ready for a merge.