ablation added

README.md

@@ -12,22 +12,30 @@ Raghunathan Ramakrishnan, Pavlo O. Dral, Matthias Rupp & O. Anatole von Lilienfe
 
 In this work, we used the QM9 dataset (that contains about 134K molecules) to demonstrate the power of EHVI-based algorithms. In the interest of time and computational cost, we used randomly selected 1000 structures. We designed a few multi-objective tasks for this project, which are not necessarily motivated by any scientific design goals. But our results show the promise of using BO for similar kind of multi-objective molecular properties optimization. For featurization, we used the coulomb matrix (CM) and smooth overlap of atomic orbitals (SOAP) methods. For GP kernels, we used RBF, and Matern kernel. We compared three different acquisition functions, namely, 1. qEHVI, 2. qNEHVI, and 3. random data acquisition. In all the simulation results provided below, we used 50 initial random seed points. One main limitation of this work is, due to time constraints, we didn't run the simulation several times to have statistically robust results.
 
-## Maximizing mu (dielectric constant) and alpha (polarizability)
+### Effect of featurizer
+
+We tried SOAP and Coulomb matrix (CM) to featurize the molecules. In our study, all things being equal, SOAP outformed CM, owing to the fact that we have to pad the CM with zero's to have the same fingerprint length. As a result, we are introducing spurious features that does not contain any signal and furhter amplifies the noise present in the data.
+
+### Effect of kernel
+
+We used RBF and Matern kernel. In this study, we found that they perform almost equally, with Matern performance being slightly better.
+
+### Maximizing mu (dielectric constant) and alpha (polarizability)
 
 This problem aims to find materials with a high dielectric constant in conjunction to a high polarizability. Below figure shows the optimization profile along with the pareto fronts for all three acquistion functions.
 
 ![maximization of mu and alpha](experiments/mu_alpha/mu_alpha_results.png)
 
-# Maximizing mu and minimizing bandgap
+### Maximizing mu and minimizing bandgap
 
 This problem is trying to find the molecules with a high dielectric constant with a low bandgap (high electrical conductivity).
 
 ![Maximization of mu with minimizing gap](experiments/mu_gap/mu_gap_results.png)
 
-# Maximizing mu and minimizing Gibbs Free Energy
+### Maximizing mu and minimizing Gibbs Free Energy
 
 This problem is trying to find the most stable (low G) molecules with a high dielectric constant.
 
 ![Maximization of mu with minimizing G](experiments/mu_G/mu_G_results.png)
 
-
+From the three experiments, we found that both `qeHVI` and `qNEHVI` performed better than the random acquisition function.