Accelerating Healthcare AI: Enabling GNN in Graph Autoencoder Techniques and Making AI Teams More Successful

Across a wide range of use cases in Healthcare AI there is actually a short list of emerging ML approaches and model architectures under investigation and early deployment. These approaches typically involve autoencoders, and especially generative autoencoders, to make maximal use of diverse data sets. The autoencoders encode the available data space into a latent space, or embedding space, with which machine learning approaches can accomplish a wide range of end use cases. There is an increasing interest in Knowledge Graphs and Graph Autoencoders, and a concomitant interest in GNNs, but these approaches are quite new in this application space. Healthcare AI startups beginning to delving into these approaches but are hindered by some key bottlenecks spanning this AI market space. In this talk we review emerging graph autoencoder techniques and discuss how Graphcore is working to accelerate Healthcare AI startups by resolving bottlenecks in Time/Cost to Train and ML Expertise.