wdayang / ReLinear

An effective algorithm for training a quadratic network

Expressivity and Trainability of Quadratic Networks

| ArXiv |

This respository includes implementations of the algorithm "ReLinear" proposed in Expressivity and Trainability of Quadratic Networks. The ReLinear encourages the model to learn suitable quadratic terms gradually and adaptively in reference to the corresponding linear terms. The ReLinear method has the following two steps. First, the quadratic weights in each neuron are set to $w^g = 0, b^g = 1$ and $w^b = 0, c = 0$. Such an initialization degenerates a quadratic neuron into a conventional neuron. Second, quadratic terms are regularized in the training process. Intuitively, there are two ways of regularization: shrinking the gradients of quadratic weights and shrinking quadratic weights.

Figure 1. Illustration of the proposed training strategy.

Figure 2. The performance of a quadratic network trained using the proposed ReLinear stragetgy, with an observed improvement than the conventional network of the same structure. $(\gamma_g,\gamma_b)$, $(\alpha_g,\alpha_b)$, and $(\beta_g,\beta_b)$ are hyperparameters of ReLinear. As these hyperparameters increases from 0, the trained model transits from the conventional model to the quadratic, and the model's performance reaches the optimality.

Folders

TrainCompactQuadraticNetworksViaReLinear: this directory contains the implementation of ReLinear on a compact quadratic network. The compact quadratic network consists of compact quadratic neurons that simplify the quadratic neuron by eradicating interaction terms. The used dataset is CIFAR10.
TrainQuadraticNetworksViaReLinear+ReZero: this directory contains implementations of ReLinear+ReZero on a quadratic network. Because the quadratic network is based on the redidual connection, we can combine the proposed ReLinear with ReZero that was devised for training residual networks. The used dataset is CIFAR10.
TrainQuadraticNetworksViaReLinear: this directory contains the implementation of ReLinear on a quadratic network. The used dataset is CIFAR10.

Running Experiments

Please first go to each directory. Each directory consists of two scripts. One is about the network, and the other is the main file.

>> python TrainCompactQuadraticNetworksViaReLinear/qresnet_smaller.py           
>> python TrainQuadraticNetworksViaReLinear+ReZero/Rezero_train_56.py    
>> python TrainQuadraticNetworksViaReLinear/qtrainer_10_5.py