If your interest is in backpropogation nets (the most popular of them all), Neural Smithing: Supervised Learning in Feedforward Artificial Neural Networks (Reed, Marks, MIT Press, 1999) covers them in great detail and may be a good choice if you can only read one book. Good coverage of techniques for optimizing learning of BP nets, even covers use of Genetic Algorithms (clssically, multilayer feedforward nets are trained using an algorithm that backpropgates errors back from an output layer thru hidden layers in order to adjust connection weights in an attempt to walk a global error surface to some minimum, but other techniques are possible, including the use of genetic algorithms).
SOMs (self organizing maps) are covered by their inventor, Kohonen, in the book Self Organizing Maps (Third Edition, Springer, 2001). These simple networks are quite powerful in some use cases (ones where you don't have any idea of what a particular dataset actually means, thus learning by supervised training is not possible).
Deep learning, including deep reinforcement learning, has several applications in the fields of nanotechnology, photonics, and micro- and nanoelectronics, etc. Key areas where deep learning techniques are being applied:
Materials Discovery and Design: Deep learning models can be used to predict the properties and behavior of nanomaterials, aiding in the discovery and design of new materials with desired characteristics.
Nanoscale Imaging and Characterization: Deep learning algorithms can enhance imaging and characterization techniques used in nanotechnology and photonics. For example, deep learning can be used to improve image reconstruction, denoising, and super-resolution of nanoscale images.
Process Optimization: Deep reinforcement learning can optimize complex manufacturing processes in micro- and nanoelectronics. By training agents to make decisions and control process variables, deep reinforcement learning can improve efficiency, yield, and performance in semiconductor and other nanofabrication processes.
Photonic Device Design: Deep learning can aid in the design of photonic devices by simulating and optimizing their performance. This includes designing efficient nanoscale waveguides, photonic circuits, and optical components using deep learning-based optimizations
Device Characterization and Failure Analysis: Deep learning can assist in the characterization and failure analysis of micro- and nanoelectronic devices. By training with and analyzing large amounts of data, deep learning models can identify patterns and anomalies, aiding in the detection, diagnosis or curing of device failures.
Quantum Computing: Deep learning techniques are also being explored for applications in quantum computing is another one among possible field applications.
It's important to get into the deep learning in these fields that is an active area of research and development, and ongoing advancements.
Deep learning (DL) has many applications in material discovery and design. For example, paper [1] retrieves subwavelength dimensions from solely far-field measurements + address inverse problem (i.e. obtaining a geometry for a desired electromagnetic response). Meanwhile, Paper [2] uses DL for forward and inverse design of nanoantenna. Besides, deep reinforcement learning (DRL) also has applications on quantum [3] and biochips optimization [4].
References
[1] Article Plasmonic nanostructure design and characterization via Deep Learning