Random masks define surprisingly effective sparse neural network models, as has been shown empirically. The resulting sparse networks can often compete with dense architectures and state-of-the-art lottery ticket pruning algorithms, even though they do not rely on computationally expensive prune-train iterations and can be drawn initially without significant computational overhead. We offer a theoretical explanation of how random masks can approximate arbitrary target networks if they are wider by a logarithmic factor in the inverse sparsity $1 / \log(1/\text{sparsity})$. This overparameterization factor is necessary at least for 3-layer random networks, which elucidates the observed degrading performance of random networks at higher sparsity. At moderate to high sparsity levels, however, our results imply that sparser networks are contained within random source networks so that any dense-to-sparse training scheme can be turned into a computationally more efficient sparse-to-sparse one by constraining the search to a fixed random mask. We demonstrate the feasibility of this approach in experiments for different pruning methods and propose particularly effective choices of initial layer-wise sparsity ratios of the random source network. As a special case, we show theoretically and experimentally that random source networks also contain strong lottery tickets.

In this paper, an algorithm is proposed for the classification of faults in three phase self-excited induction generators using deep neural networks, on the basis of their voltage and current waveforms. Three phase self-exited induction generators, which are mostly used in wind power stations, are often connected to the national grid. Therefore, the transient stability analysis of this machine, prior and post symmetrical and unsymmetrical short circuit faults is one of the main concerns in power system security and operation. In this study, voltage and current waveforms of faults have been simulated in the Simulink environment, for different conditions of fault. Following this, visual time-frequency representations of the fault signals called scalograms are created, using continuous wavelet transform. Finally, a deep convolutional neural network is used for classification of the fault signals. Experimental results show that a final accuracy of 82.14% as well as real-time inference is achieved on the validation set, using the proposed scheme.