WaveNet
- 1. WaveNet: A Generative Model for Raw Audio TIS + Albert 茶氏 2017/01/24 恷貧肪伏 tsuguo_mogami@albert2005.co.jp
- 2. Why? ? Autoregressive model (e.g. PixelCNN)が措く撹孔している。 ? ? ★ 咄蕗はどうであろうか。 ? それをRNNより紳糞弔CNNで佩いたい。
- 3. Contributions ? これまでにない瞳|の咄蕗栽撹。 ? Dilated convolutionを聞い、寄きな鞭否勸を隔つにもvわらず 紳糞弔淵〒`キテクチャ ? 咄蕗JRも
- 5. stack of dilated causal convolutional layers 鞭否勸の寄の古廷蹐任△蝓gHはResNetLblockのRり卦しです
- 6. Repetition Structure 1,2,4,´,512, 1,2,4,´,512, 1,2,4,´,512. Suspected to be repeating the 1´512 blocks 16 times
- 8. residual block and the entire architecture ちょっとわかりにくいので噸宥の燕に宙きなおします
- 9. ???
- 10. Gated activation units ? ? K: layerindex, f for filter, g for gate ? : elementwise multipulication ? h: condition (person, text, etc.) ? Why? ? PixelCNN (1606.05328)で秘 ? それ參念のCNN伏撹モデルがPixelRNNに喪ったのは LSTMのゲ`ト夛のせいだと深えてLSTM貌のゲ` トを秘した
- 12. Things not described and Guesses ? Kernel size of the dilation filters 2 ? Number of the layers (ResNet-blocks) 4*10~ 6*10 ? Number of the channels in hidden layers hundreds? 256? ? the other activation function in a Res-block? may be no ? Batch normalization no reason not to use ? Sampling frequency ｀at least 16kHz¨ ? Where to let the skip connection out? Every 10? ? Skip connections have weights yes?
- 13. Experiments
- 14. Text-to-Speech (TTS) ? Single-speaker speech dataset ? North American English dataset: 24.6hr ? Mandarin Chinese dataset: 34.8hr ? Receptive field 240ms ? Ad hoc architecture as ★ WaveNet Audio(t) Yet another model Liguistic feature h_i (possibly phoneme) Another model Fundamental frequency F0(t) duration(t) Liguistic feature h(t) ☆猟とは`った催を聞っています。
- 15. TTS: Mean Opinion Score https://deepmind.com/blog/wavenet-generative-model-raw-audio/
- 16. Speech Recoginition ? TIMIT dataset (possibly ~4hrs) ? Add pooling layer after dilated convolution ? of 160x down sampling (Does it mean 7th layer?) ? Then a few non-causal convolutions. ? Loss to predict the next sample (same as ordinary WaveNet) ? And a loss to classify the frame ? 18.8PER, which is best score among raw-audio models.
- 17. End
- 18. (Multi-speaker) Speech Generation ? Conditioned on the speaker ? 44 hours of data (from 109 speakers)
- 19. TTS: Mean opinion score
- 20. μ-law transformation (ITU-T, 1988) ? ? で-1,1のgを256蛍護している。 ? 寄j委には log でコ`ドしているだけ。