AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is There Hope for Machine Learning?
The training set is awful. The people who worked on it and data that I get, often they're dead. And the files that I've used, some of them were hand drawn by someone tired in the 70s. So my next step is to improve this training set by hand and then try to see if I can apply machine learning on the full code of the full 16,000 molecules and improve them all. But really, I need to be able to test the outcomes with experimental data,. which means convincing someone in a lab to spend a lot of money putting very dangerous gases in chambers and measuring them at outrageous temperatures.