OptNet: Differentiable Optimization as a Layer in Neural Networks — mentioned by dgacmu

Wed, 15 Nov 2017 02:32:29 GMT

"just a bunch of weights and layers and parameters" -- I think you and the GP are agreeing. That's the definition of standard: If the model can be expressed using the currently-blessed set of layer definitions in CoreML, then yes. But if you're doing nonstandard stuff with weird control flow behavior, or RNNs that don't map into some of the common flavors, then all bets are off. An example: Some of my colleagues put a QP solver in tandem with a DNN, so that the neural network could 'shell out' to the solver as part of its learning, and learned to solve small sudoku problems from examples…

Paper and discussion history. A mention is not an endorsement.

OptNet: Differentiable Optimization as a Layer in Neural Networks — mentioned by shriphani

Mon, 12 Jun 2017 17:33:47 GMT

Indeed. A simple solution is to add a differential optimization layer (see Amos et al): https://arxiv.org/abs/1703.00443

Paper and discussion history. A mention is not an endorsement.

OptNet: Differentiable Optimization as a Layer in Neural Networks · HackerPapers discussions

OptNet: Differentiable Optimization as a Layer in Neural Networks — mentioned by dgacmu

OptNet: Differentiable Optimization as a Layer in Neural Networks — mentioned by shriphani