Variation Auto-encoder (VAE) NN input NN Encoder → output Decoder code VAE Minimize reconstruction error NN input NN Encoder exp output Decoder C3 ci exp(oi)xei +mi From a normal Minimize 3 distribution (exp(o)-(1+o)+(m)2) Ref:Auto-Encoding Variational Bayes, i=1 https://arxiv.org/abs/1312.6114 门义入丁川工
电子科技大学研究生《机器学习》 NN Encoder NN Decoder code input output VAE NN Encoder input NN Decoder output m1 m2 m3 From a normal distribution X + Minimize reconstruction error exp Minimize Variation Auto-encoder (VAE) Ref: Auto-Encoding Variational Bayes, https://arxiv.org/abs/1312.6114
VAE-Writing Poetry 56 NN sentence NN exp sentence Encoder Decoder code Code Space i went to the store to buy somsmnceaqsuy some groceries. i were to buy any groceries. "come with me,she said. "talk to me,"she saididon't worry about it, she said. Ref:http://www.wired.co.uk/article/google-artificial-intelligence-poetry Samuel R.Bowman,Luke Vilnis,Oriol Vinyals,Andrew M.Dai,Rafal ofouies-Sam Bengio Generating Sentenees学释缓哭y%生 arYiu nronring 2015 《机器学习》
电子科技大学研究生《机器学习》 Ref: http://www.wired.co.uk/article/google-artificial-intelligence-poetry Samuel R. Bowman, Luke Vilnis, Oriol Vinyals, Andrew M. Dai, Rafal Jozefowicz, Samy Bengio, Generating Sentences from a Continuous Space, arXiv prepring, 2015 VAE - Writing Poetry NN Encoder sentence NN Decoder sentence code Code Space i went to the store to buy some groceries. "come with me, " she said. i store to buy some groceries. i were to buy any groceries. "talk to me, " she said".don’t worry about it, " she said. … …
0800s 电子科技大学研究生《机器学习》
电子科技大学研究生《机器学习》
Problems of vAE 956 It does not really try to simulate real images code NN As close as Decoder Output possible One pixel difference One pixel difference from the target from the target Realistic Fake 电子科技大学研究生《机器学习》
电子科技大学研究生《机器学习》 Problems of VAE o It does not really try to simulate real images NN Decoder code Output As close as possible One pixel dif erence from the target One pixel dif erence from the target Realistic Fake