Fig. 1. Deep neural networks used to predict output energy of main amplifier^[12]. (a) Internal optical circuit and main modules of main amplifier; (b) relationships between input and output energy of beam; (c) prediction result and fitting method

Fig. 2. Beam shaping using diffractive neural networks^[15]. (a) Principle diagram of diffractive neural network; (b) simulation results of beam shaping by DOE

Fig. 3. BiLSTM-CNN model used for fiber fault diagnosis^[20]

Fig. 4. Surface roughness prediction^[24]. (a) Measured roughness values of laser-cut sample versus sample thickness; (b) prediction error of roughness versus size of training dataset

Fig. 5. Neural network model and surface topography analysis^[32]. (a) Topological network structure; (b) macroscopic image; (c) comparison of polished and unpolished surfaces; (d) 3D topographic image of polished surface; (e) cross-section microstructure of sample after polishing; (f) microstructures of heat affect zone and polished layer; (g) nanoindentation load-displacement curves

Fig. 6. Laser bone drilling experiment^[42]. (a) Feature identification and ablation control; (b) spectral amplitude at focus position versus time; (c) spectral amplitude at defocusing position versus time

Fig. 7. Surface microstructures^[43]. (a) Training error of neural network model; (b) images of various surface microstructures

Fig. 8. Relationship between weld pool and keyhole constructed by different neural networks^[46]. (a) Feature extraction of (a) melting pool and (b) keyhole; (c) relationship between weld pool and keyhole constructed by radial basis function neural network; (d) relationship between weld pool and keyhole constructed by BP neural network; (e) relationship between weld pool and keyhole constructed by generalized regression neural network; (f) relationship between weld pool and keyhole constructed by evolutionary neural network

Fig. 9. Schematics of machine learning (ML) assisted composition design of Fe-Ni-Ti-Al novel maraging steel (NMS)^[55]. (a) Feature selection; (b) data collection; (c) ML by various algorithms; (d) composition optimization of alloy elements; (e) time-dependent dynamic precipitation behavior; (f) powder morphology and elemental mapping