I think that gasnet neural networks that took into account the locally inhibitive? impact of nitrous oxide are far superior to traditional ANN. But I just read that recently, so maybe misunderstood that.