1. dynamic_yoho

Yoho with spliter and combiner

1.1. Content

1. dynamic_yoho

1.2. Getting Started

Python Package:

pip install ipykernel torch librosa sklearn

Cuda install:

follow the official CUDA Download to get started!
Using CPU:

CPU will be used in this repo instead of CUDA, using git diff to check the change in distributed_model.ipynb

1.3. Description

1.3.1. IA-Net Module

In comparison to previous the TS Method, our IA-Net aims to extract the abstract anomalous features of each source from the mixed signal X to replace the reconstructed signal in AutoEncoder.

Here is the structure of IA-Net Module:

1.3.1.1. GLayerNorm

Extends from nn.Module, mainly work is normalize the simple.

  # [N, C, T] -> [N, T, C]
  sample = torch.transpose(sample, 1, 2)
  # Mean and variance [N, 1, 1]
  mean = torch.mean(sample, (1, 2), keepdim=True)
  var = torch.mean((sample - mean) ** 2, (1, 2), keepdim=True)
  sample = (sample - mean) / torch.sqrt(var + self.eps) * \
  self.gamma + self.beta
  # [N, T, C] -> [N, C, T]
  sample = torch.transpose(sample, 1, 2)
  return sample

1.3.1.2. Bottleneck

Encoder

size calculate can base on this formula:

Naibaowjk/dynamic_yoho