2002, when Ronan Collobert, Samy Bengio and Johnny Mariéthoz at IDIAP described the C++ library. The better-known Lua-based Torch7 followed in 2011.
Card 1 of 259. Answer: 2002, when Ronan Collobert, Samy Bengio and Johnny Mariéthoz at IDIAP described the C++ library. The better-known Lua-based Torch7 followed in 2011.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
Torch is usually dated to 2002, when Ronan Collobert, Samy Bengio and Johnny Mariéthoz at the IDIAP research institute in Martigny, Switzerland, described it in a research report, "Torch: a modular machine learning software library". It was an open-source C++ library of machine learning algorithms; a first version had been announced on the Connectionists mailing list in September 2001.
Early Torch went well beyond neural networks: multilayer perceptrons, mixtures of experts, support vector machines, Gaussian mixture models and hidden Markov models. The Torch most people remember is Torch7, described in 2011 by Collobert, Koray Kavukcuoglu and Clément Farabet. It wrapped fast C and CUDA code in the Lua scripting language and became a main research tool at DeepMind, Facebook and Twitter.
Its decline is a common follow-up. DeepMind switched to TensorFlow in April 2016, and Facebook's researchers kept Torch's C backend but replaced Lua with Python, releasing PyTorch in January 2017. Lua Torch is no longer developed, so torch in modern code means PyTorch.
