Fig. 4(b) shows the schematic of an output neuron. As can be seen, it is similar to the hidden neuron, but it uses a RAM module (instead of a ROM) with the aim of enabling ELM adaptation; distributed RAM memories are used to provide faster access to the data [15]. Moreover, the output neuron in an ELM neural network is slightly simpler than the hidden neuron because it does not need a bias register. In addition, the sum of products of these neurons is directly driven to the corresponding output without passing through an activation function. Both, ROM and RAM memories in Fig. 4 are implemented as a part of the neurons (i.e. in the same component). As a consequence, the path between the output of the memories and the input of the neurons’ DSP is as short as possible, thus decreasing the propagation delay of the signals.