Waymo collaborates with Google’s DeepMind to train neural networks of self-driving cars

DeepMind has developed a new training method called Population Based Training (PBT) which allows an experimenter to quickly choose the best set of hyperparameters and model for the task

Waymo has teamed up with Google’s DeepMind to train neural networks for self-driving cars, the company said in a blog posted on Medium. According to Waymo, its self-driving cars employ neural networks to perform several driving tasks, ranging from detecting objects and predicting how others will behave to planning a car’s next move. However, training an individual neural net usually require weeks of fine tuning and experimentation and large amount of computational power.