# Paper review: ALL YOU NEED IS A GOOD INIT

This work proposes yet another way to initialize your network, namely LUV (Layer-sequential Unit-variance) targeting especially deep networks.  The idea relies on lately served Orthogonal initialization and fine-tuning the weights by the data to have variance of 1 for each layer output.

The scheme follows three stages;

1.  Initialize weights by unit variance Gaussian
2.  Find components of these weights using SVD
3.  Replace the weights with these components
4.  By using minibatches of data, try to rescale weights to have variance of 1 for each layer. This iterative procedure is described as below pseudo code.

In order to describe the code in words, for each iteration we give a new mini-batch and compute the output variance. We compare the computed variance by the threshold we defined as $Tol_{var}$ to the target variance 1.   If number of iterations is below the maximum number iterations or the difference is above $Tol_{var}$ we rescale the layer weights by the squared variance of the minibatch.  After initializing this layer go on to the next layer.

In essence, what this method does. First, we start with a normal Gaussian initialization which we know that it is not enough for deep networks. Orthogonalization stage, decorrelates the weights so that each unit of the layer starts to learn from particularly different point in the space. At the final stage, LUV iterations rescale the weights and keep the back and forth propagated signals close to a useful variance against vanishing or exploding gradient problem , similar to Batch Normalization but without computational load.  Nevertheless, as also they points, LUV is not interchangeable with BN for especially large datasets like ImageNet. Still, I'd like to see a comparison with LUV vs BN but it is not done or not written to paper (Edit by the Author: Figure 3 on the paper has CIFAR comparison of BN and LUV and ImageNet results are posted on https://github.com/ducha-aiki/caffenet-benchmark).

The good side of this method is it works, for at least for my experiments made on ImageNet with different architectures. It is also not too much hurdle to code, if you already have Orthogonal initialization on the hand. Even, if you don't have it, you can start with a Gaussian initialization scheme and skip Orthogonalization stage and directly use LUV iterations. It still works with slight decrease of performance.

• Dmytro Mishkin

Hi, I am the author of the paper.
We have comparison with BN in paper on CIFAR-10 - see Figure 3.
As for Imagenet, it was too long to test on it, but now it is available here - https://github.com/ducha-aiki/caffenet-benchmark see activations (everything was initialized with LSUV) and bn-activations section.