Initially, every node in the graph is represented as a vector, using any preferred featurization — atom identity, atom charge, etc. Then, in a series of message passing steps, every node broadcasts its current vector value to each of its neighbors. An update function then takes the collection of vectors sent to it, and generates an updated vector value. This process can be repeated many times, until finally all of the nodes in the graph are summarized into a single vector via summing or averaging. That single vector, representing the entire molecule, can then be passed into a fully connected network as a learned molecular featurization. This network outputs a prediction for odor descriptors, as provided by perfume experts.
The idea is that the CPU spawns a thread per element, and the GPU then executes those threads. Not all of the thousands or millions of threads actually run in parallel, but many do. Specifically, an NVIDIA GPU contains several largely independent processors called "Streaming Multiprocessors" (SMs), each SM hosts several "cores", and each "core" runs a thread. For instance, Fermi has up to 16 SMs with 32 cores per SM – so up to 512 threads can run in parallel.
Bookmarks on this page are managed by an admin user. 