NLP related research (off-topic)

[nicola]

GAN (can we come up with a discriminator that can't differentiate machine from humans?)

• ⭐️ Generative Adversarial Networks
• Improved Techniques for Training GANs
• InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
• How not to train Generative models
• Turing test with turing machines
• Building machine that learn and think like people
• Visual turing test

Decentralize learning (can we preserve the inputs during a computation?)

• NIPS workshop
• SSGD
• CryptoNets
  • problem: hospital wants to use cloud service, but can't due to legal compliance
  • convert learned neural networks to NN that can be applied on encrypted data
  • assumption: model already created safely
  • leveled homomorphic encryption scheme (Bos et al 2013): adding multiplying, but requires the complexity of the arithmetic circuit
  • use different tricks in neural networks,