以前介绍过一个backtrader的参数优化包optunity,此包有个缺点,假设参数是列表[5,6],optunit会搜索5和6之间的小数,也就是它是连续搜索,而不是离散搜索,这导致很多无效搜索,因为我们实际只想搜索5和6这两个数。最近发现一个包可以执行离散搜索,那就是gradient_free_optimizers。以下案例内容来自扫地僧backtrader技术教...
Simple and reliable optimization with local, global, population-based and sequential techniques in numerical discrete search spaces. - SimonBlanke/Gradient-Free-Optimizers
Noise-free, good quality (SNR>100) datasets were inverted usingLevenberg–Marquardt: LM (gradient) and Nelder–Mead: NM (gradient-free) methods. All simulations were performed using acoupled simulation-optimization model. Performance of the two optimizers is evaluated by comparing model predictions ...
---> 33 opt.search(model, n_iter=5) D:...\gradient_free_optimizers\search.py in search(self, objective_function, n_iter, max_time, max_score, memory, memory_warm_start, verbosity, random_state, nth_process) 119 self.nth_process = nth_process 120 --> 121 self._init_search() 122...