Another breakthrough for autonomous driving: Random Contrast Learning, a new CPU-based algorithm, performs much better than a neural network in autonomous simulators. The experiments: • A controlled simulation environment • An autonomous vehicle with an array of 8 sensors • A race track • Two side-by-side algorithms • 100 trials The algorithms were LuminaRCL (proprietary implementation of Random Contrast Learning) and an artificial neural network. The vehicle gradually learned to navigate the track using both algorithms. The Random Contrast Learning implementation completed a full lap with an 81% success rate after 15,000 training steps. The neural network required 150,000 steps to achieve a paltry 23% success rate. Here is more information about the experiments: bit.ly/RCLvsNN-CarSim… Thanks to the Lumina team for collaborating with me on this post and helping me understand PrismRCL, their desktop application you can use to run their Random Contrast Learning algorithm. You can try it out for free by going to this link: lumina247.com/prismrcl-sign-…