A Neural Spiking Approach Compared to Deep Feedforward Networks on Stepwise Pixel Erasement
This work addresses robustness in object recognition for partially occluded scenarios, but it is incremental as it compares a spiking network to existing methods on a standard dataset.
The paper tackled object recognition under partial occlusion by comparing a two-layer spiking neural network trained with a biologically plausible rule to deep convolutional networks on MNIST with stepwise pixel erasure, finding that the spiking approach achieves good accuracy and robustness.
In real world scenarios, objects are often partially occluded. This requires a robustness for object recognition against these perturbations. Convolutional networks have shown good performances in classification tasks. The learned convolutional filters seem similar to receptive fields of simple cells found in the primary visual cortex. Alternatively, spiking neural networks are more biological plausible. We developed a two layer spiking network, trained on natural scenes with a biologically plausible learning rule. It is compared to two deep convolutional neural networks using a classification task of stepwise pixel erasement on MNIST. In comparison to these networks the spiking approach achieves good accuracy and robustness.