Neural network learning optimization method based on particle swarm optimization algorithm

Abstract

The invention discloses a method of learning and optimizing a neural network based on a particle swarm optimization algorithm. The method comprises the steps: choosing macroscopic directional parameters of the particle swarm optimization algorithm and representing particle formation and swarm size; setting microscopic directional parameters of the particle swarm optimization algorithm and updating the own motions of particles according to social search and cognitive search; combining with the present parameters, training and learning with the particle swarm optimization algorithm and enabling the particles in the particle swarm to continuously approach an optimal particle; performing time estimation for rendering data by utilizing a trained neutral network model, and formulating swarm working and dispatching strategies to achieve a purpose of reducing actual rendering time. The method of learning and optimizing the neural network based on the particle swarm optimization algorithm is convenient to calculate, has rapid solution speed and is suitable for solving real numbers; and meanwhile, the easy convergence of the BP neural network to the locally optimal solution can be avoided.

Description

technical field [0001] The invention relates to the field of graphics realistic rendering, in particular to a neural network learning and optimization method based on a particle swarm optimization algorithm. Background technique [0002] Photorealistic rendering is a very important part of digital film and television production, and it is also the most time-consuming part. Since it takes a very long time to render a high-quality work, it is very important to submit tasks to the cluster for parallel rendering. In order to improve the utilization efficiency of the cluster, it is necessary to formulate a reasonable cluster job scheduling strategy, and the scheduling strategy is related to the rendering Time is closely related. If the image rendering time can be predicted, it will be very helpful for the formulation of scheduling strategies. [0003] At this stage, resource estimation methods are mainly divided into two types: resource estimation based on hard-coded methods and...

AI Technical Summary

Problems solved by technology

[0010] ①Local minimization problem: From a mathematical point of view, the traditional BP neural network is a local search optimization method that solves a complex nonlinear problem. The weight of the network is gradually improved and adjusted in the local direction, which will It will cause the local optimal weight to converge to the local minimum point, resulting in network training failure

[0011] ②The convergence speed of the network is too slow: Since the BP neural network adopts the gradient descent algorithm to train the network, the function it optimizes is relatively complicated. Therefore, the "zigzag" phenomenon is inevitable, that is, if the vicinity of the minimum value is relatively flat, the algorithm will be in the Staying near the minimum value for a long time, the convergence is slow, which makes the BP algorithm inefficient; at the same time, because the optimized function is very complex, it will appear a flat area when the output of the neuron is close to the result, and the output in this area changes very little Therefore, the error change is also very small, but the training is still going on, but the process seems to be stopped; in the BP neural network model, the weight update rules are assigned to the neural network in advance, and the update value of each iteration cannot be calculated, so is also less efficient

Images