PREDICTION OF 12 PHOTONIC CRYSTAL FIBER OPTICAL PROPERTIES USING MLP IN DEEP LEARNING
Abstract : Photonic crystal fibers (PCFs) are the specialized optical waveguides that led to many interesting applications ranging from nonlinear optical signal processing to high-power fiber amplifiers. In this paper, machine learning techniques are used to compute various optical properties including effective index, effective mode area, dispersion and confinement loss for a solid-core PCF. These machine learning algorithms based on artificial neural networks are able to make accurate predictions of above mentioned optical properties for usual parameter space of wavelength ranging from 0.5-1.8 µm, pitch from 0.8-2.0 µm, diameter by pitch from 0.6-0.9 and number of rings as 4 or 5 in a silica solid-core PCF. We demonstrate the use of simple and fast-training feed-forward artificial neural networks that predicts the output for unknown device parameters faster than conventional numerical simulation techniques Computation runtimes required with neural networks (for training and testing) and Lumerical Mode Solutions are also compared.
? Photonic crystal fiber was first proposed by Knight et al.
? in 1996, which consists of a core with the periodic arrangement of air holes running along the length of the fiber. The core of the PCF can be solid or hollow
? proposed a deep learning-based algorithm using the dimensionality reduction technique to understand the properties of electromagnetic wave-matter interaction in nanostructures.
? A geometric deep learning approach has also been reported to study nanophotonics structures In 2018, extreme learning machine and deep learning were used for computing dispersion relations and optimization of Q-factors for photonic crystals
? The accuracy of the model depends upon how well the dataset is aligned to the problem to be solved. Variant PCFs were simulated by changing some geometric property values and their optical properties were calculated.
? Choosing the above-mentioned parameters are based on our prior experience of similar problems and a thorough study has been carried out in our previous work
? We could increase the epochs to a higher value but this would increase the simulation time, and may also lead to the overfitting problem
? We demonstrate the use of simple and fast-training feed-forward artificial neural networks that predicts the output for unknown device parameters faster than conventional numerical simulation techniques. Computation runtimes required with neural networks (for training and testing) and Lumerical MODE solutions are also compared.? this paper, machine learning techniques are used to compute various opticalproperties including e?ective index, e?ective mode area, dispersion and con?nement loss for asolid-core PCF.
? In 2018, extreme learning machine and deep learning were used for computing dispersionrelations and optimization of Q-factors for photonic crystals.
? The ANN/MLP architecture parameters are introduced in this section along with the PCF typethat is used for generating the dataset. The ?rst step of the training procedure of an ANN model is to have a ?nite and appropriate labeled dataset.
? An ANN/MLP model with 3 hidden layers and 50 nodes/neurons in each layer was usedthroughout this paper, as shown in Fig. 2. These hidden layers were fully interconnected, whichmeans each node/neuron of a layer is connected to each node/neuron in the following layer.
|