Kyunghyun Cho

Kyunghyun Cho

A South Korean researcher, now a professor at New York University, who co-authored the 2014 encoder-decoder paper that introduced the Gated Recurrent Unit, a simpler alternative to LSTM, and co-authored the attention mechanism with Dzmitry Bahdanau and Yoshua Bengio the same year. That pair of papers defined the neural machine translation architecture that preceded the Transformer. He is also a persistent voice for careful evaluation and against overclaiming in the field. (See also: Dzmitry Bahdanau, Yoshua Bengio, Transformer, Benchmark)