Quantum Gaussian Mixture Models - Week 9

This week, I've researched the impact of lambda and phi modeling. I did several experiments, got some insights from the researches, organized them in some documents. About lambda, although the higher lambda is, the more QGMM can be constrained, it's trembled on the convergence side when finding the center of the observations, so it is somewhat important for finding an proper lambda depending the data sets. With regards to phi modeling, when I changed phi to a trainable variable, the training process was more stable than before. I think it's because we can get the proper range of phi.

Thanks for reading :)