{փWv

Vision

At the head of applications of CVNNs in engineering, we have coherent systems, in particular, sensing and imaging systems. Our daily life widely depends on radars and sonars where we use coherent waves to probe the world. Conventional systems treat information obtained by such systems just as they do for other types of image data. However, we can develop various new functions by processing the image by appropriate CVNNs utilizing the fact that the information arose from wave phenomena.

In the future, the CVNNs will become the most fundamental framework to realize quantum-wave electronic devices having learning and/or self-organizing dynamics. VLSI technology advances more miniaturization. Then we face to crystal inhomogeneity and temporal instability in the bit processing, i.e., the difficulty in holding g1h or g0h stably. In such a stochastic world, the neural networks make a significant contribution to the electronic devices as the most fundamental information-processing framework. Ultimately, in addition, such devices deal with electron-wave, maybe with its spin. We therefore require phase sensitive learning and self-organizing dynamics, i.e., the CVNNs. Related fields includes super-conducting devices, quantum neural devices, and learning optical computing.

In computer graphics and virtual reality, we inevitably need quaternion neural networks as we mentioned the background. In deep space missions, the quaternion networks is indispensable to control thrust and robot arms since a command leaving the earth requires long period before it reaches the spaceship. We need a self-reliant system based on quaternion neural networks.

In addition, the phase information increases its significance even in physiology to elucidate the interaction between macroscopic potential wave and microscopic neuron firing. An example is the relationship between the neural spike timing with reference to the theta wave and the short-term memory.

Figure 1 summarizes possible application fields and their basis [A.Hirose, gComplex-Valued Neural Networks,h Springer (2006)].