The adaptation is performed using a linear regression transformation matrix based on an NMF framework.
This method requires only a small amount of parallel data, where a linear regression transformation matrix is used to adapt a source dictionary to a target dictionary and it is estimated in an NMF framework.
In VC, the source dictionary is constructed using sufficient source speaker data, and it is adapted using a small amount of parallel data (about ten words only) in order to obtain the target dictionary, where a linear regression transformation matrix (affine matrix) is trained based on NMF.
Raw data were first imported into a Genetraffic duo database (Iobion Informatics, Toronto, Canada), local background-subtracted and normalized using a Lowess (locally weighted linear regression) transformation.
The second design we assess for comparison is based on Bayesian isotonic regression transformation (BIT) [ 23].
As in the previous section, we use Bayesian isotonic regression transformation to compute and monitor toxicity continuously.
Next, to borrow strength (i.e., the ordering constraint) across the two dose levels, we apply a Bayesian isotonic regression transformation approach [ 23].
MaxED: Maximum effective dose; MTA: Molecularly targeted agent; MTD: Maximum tolerated dose; LOXL2: Lysyl oxidase homolog 2; DLT: Dose-limiting toxicity; BHT: Bayesian hypothesis testing; BMA: Bayesian model averaging; BIT: Bayesian isotonic regression transformation.
Our simulation results suggest that the BHT-A design performs better overall than the BMA design, the independent single-arm design using Bayesian hypothesis tests with a nonlocal alternative prior, and the Bayesian isotonic regression transformation (BIT -based design.
A nonlinear regression model is derived from the linear regression model through regression transformation listed as follows: (1) y i = β 0 + β 1 x 1 + β 2 x 2 + ⋯ + β p x p + ε i, ε i ∈ N 0, σ.
