Inputs are very first handed by way of some completely related layer, to your double-layer residual multihead focus as shown in Fig. 7. Residual networks (Kaiming He, 2016), include feedforward to avoid neurons from encountering exploding or vanishing gradients all through the educational process. The thoroughly connected levels within the residual