Validity Improvement in MolGAN-Based Molecular Generation

Designing molecules that have desired properties is one of the challenging tasks drug design. Among many molecular generative models, a adversarial network (GAN), able to generate molecule structures with desirable chemical via reinforcement learning. Generating valid foremost task any model, since invalid cannot be synthesized. We base our research on (MolGAN) architecture investigate how validity score influenced in different scenarios. First, we verify Vanilla GAN structure can produce measure, and reward network, along GAN, further increase learning manner. Then, procedure for solely optimizing tested, followed by an assessment maintenance while other are being optimized. found multiple aspects, including loss functions, hyper parameters, training sequences, must carefully considered optimized raise generation alone or concurrence property scores.