Algorithm 1: Model transfer algorithm based on an autoencoder
	Input: The master spectrometer data X, slave spectrometer data Y;
	Output: The reconstructed data after model transfer from the instrument;
1	Main procedure
2	Bias B and weight W in the autoencoder neural network of the master and slave instrument are randomly initialized. The parameters such as training times and learning rate of the model are set;
3	Compute the hidden layer variables $H_x$ and $H_y$, calculate the optimization objective function and optimize the whole model;
4	Optimized the hyperparameters of the autoencoder according to the Bayesian optimization algorithm, and the bias B and weight W were updated continuously by the gradient descent method until the optimal objective function value was found;
5	Save the slave encoder and the master decoder models to form the model transfer structure based on the improved autoencoder.