torch_specinv.metrics

torch_specinv.metrics.sc(input, target)

The Spectral Convergence score is calculated as follow:

\[\mathcal{C}(\mathbf{\hat{S}}, \mathbf{S})=\frac{\|\mathbf{\hat{S}}-\mathbf{S}\|_{Fro}}{\|\mathbf{S}\|_{Fro}}\]

Returns:scalar output in db scale.

torch_specinv.metrics.ser(input, target)

The Signal-to-Error Ratio (SER) is calculated as follow:

\[SER(\mathbf{\hat{S}}, \mathbf{S})= 10\log_{10}\frac{\sum \hat{s}_i^2}{\sum (\hat{s}_i - s_i)^2}\]

Returns:scalar output.

torch_specinv.metrics.snr(input, target)

The Signal-to-Noise Ratio (SNR) is calculated as follow:

\[SNR(\mathbf{\hat{S}}, \mathbf{S})= 10\log_{10}\frac{1}{\sum (\frac{\hat{s}_i}{\|\mathbf{\hat{S}}\|_{Fro}} - \frac{s_i}{\|\mathbf{S}\|_{Fro}})^2}\]

Returns:scalar output.