A Semi-supervised Classification Using Gated Linear Model

A semisupervised classifier based on piecewise linear regression model using gated linear network

“…It is well known that a neural network can be trained to capture data manifold , and the captured data manifold can guide the generation of gate control signals in our piecewise linear regression model . In Section 3, we design a semisupervised gating mechanism based on the neural network, which can capture data manifold from both labeled and unlabeled data, to generate gate control signals.…”

“…Instead of using an unsupervised manner , the gating mechanism is trained as a semisupervised neural network that can capture the data manifold from both labeled and unlabeled data by using a k ‐sparse aggressive sparsity strategy. In the second step, the gated linear network is first transformed into a linear regression form, and the linear parameters are then optimized globally by an LapRLS algorithm using a kernel function comprising the gating sequences obtained in the first step . Moreover, we use the kernel function as a similarity measurement to construct the graph in LapRLS.…”