Normalization of Neural Networks using Analytic Variance Propagation
نویسندگان
چکیده
We address the problem of estimating statistics of hidden units in a neural network using a method of analytic moment propagation. These statistics are useful for approximate whitening of the inputs in front of saturating non-linearities such as a sigmoid function. This is important for initialization of training and for reducing the accumulated scale and bias dependencies (compensating covariate shift), which presumably eases the learning. In batch normalization, which is currently a very widely applied technique, sample estimates of statistics of hidden units over a batch are used. The proposed estimation uses an analytic propagation of mean and variance of the training set through the network. The result depends on the network structure and its current weights but not on the specific batch input. The estimates are suitable for initialization and normalization, efficient to compute and independent of the batch size. The experimental verification well supports these claims. However, the method does not share the generalization properties of BN, to which our experiments give some additional insight.
منابع مشابه
Efficient Short-Term Electricity Load Forecasting Using Recurrent Neural Networks
Short term load forecasting (STLF) plays an important role in the economic and reliable operation ofpower systems. Electric load demand has a complex profile with many multivariable and nonlineardependencies. In this study, recurrent neural network (RNN) architecture is presented for STLF. Theproposed model is capable of forecasting next 24-hour load profile. The main feature in this networkis ...
متن کاملFeed-forward Uncertainty Propagation in Belief and Neural Networks
We propose a feed-forward inference method applicable to belief and neural networks. In a belief network, the method estimates an approximate factorized posterior of all hidden units given the input. In neural networks the method propagates uncertainty of the input through all the layers. In neural networks with injected noise, the method analytically takes into account uncertainties resulting ...
متن کاملOn the use of back propagation and radial basis function neural networks in surface roughness prediction
Various artificial neural networks types are examined and compared for the prediction of surface roughness in manufacturing technology. The aim of the study is to evaluate different kinds of neural networks and observe their performance and applicability on the same problem. More specifically, feed-forward artificial neural networks are trained with three different back propagation algorithms, ...
متن کاملCosine Normalization: Using Cosine Similarity Instead of Dot Product in Neural Networks
Traditionally, multi-layer neural networks use dot product between the output vector of previous layer and the incoming weight vector as the input to activation function. The result of dot product is unbounded, thus increases the risk of large variance. Large variance of neuron makes the model sensitive to the change of input distribution, thus results in poor generalization, and aggravates the...
متن کاملClassification of ECG signals using Hermite functions and MLP neural networks
Classification of heart arrhythmia is an important step in developing devices for monitoring the health of individuals. This paper proposes a three module system for classification of electrocardiogram (ECG) beats. These modules are: denoising module, feature extraction module and a classification module. In the first module the stationary wavelet transform (SWF) is used for noise reduction of ...
متن کامل