论文部分内容阅读
提出用人工智能算法——粒子群优化算法(PSO)对CNNE模型进行训练,并针对标准粒子群算法易限于局部极小点的局限性,采用了一种带有梯度加速的粒子群算法,通过引入梯度信息来影响粒子速度的更新.为防止陷入局部最优,在群体最优信息陷入停滞时,对部分粒子进行重新初始化,从而保持群体的活性,减小群体陷入局优的可能性.采用粒子群算法训练的CNNE模型较原来的分布式最速下降法而言,在保证精度的前提下,提高了算法的收敛速度,解决了发射率的在线实时测量问题.
The artificial intelligence algorithm (PSO) is used to train the CNNE model. Aiming at the limitation of standard particle swarm optimization (PSO) easily to the local minimum point, a particle swarm optimization algorithm with gradient acceleration The gradient information is introduced to affect the update of the particle velocity.In order to avoid falling into the local optimum, some particles are reinitialized when the optimal population information stagnates, so as to maintain the activity of the population and reduce the probability of the population getting into the superiority. Compared with the original distributed steepest descent method, the CNNE model trained by particle swarm optimization improves the convergence speed of the algorithm and solves the online real-time measurement of emissivity under the premise of ensuring accuracy.