Интерполяция, аппроксимация и иже с ними (пакет alglib) - страница 12
Вы упускаете торговые возможности:
- Бесплатные приложения для трейдинга
- 8 000+ сигналов для копирования
- Экономические новости для анализа финансовых рынков
Регистрация
Вход
Вы принимаете политику сайта и условия использования
Если у вас нет учетной записи, зарегистрируйтесь
нелинейные за что обидели?
а комбинированные ядра. Задача уменьшить ошибки за счет малой крови, что у меня успешно получается. Намутите 150 слоев в нейросети а потом удаляйте по 1 слою\нейрону. Будет хуже и медленнее.думаю нелинейные вектора, на меня не обидятся))
не понял о каких операциях в нейросети речь, возможно что то типа свертки...
Sounds Great about the training speed:)
But are you sure that while trade entry the algorithm can check such large number of trained models so quickly to make an accurate trading decision?
I mean how much time do you expect on average it will take to check the trained data before trade entry or exit in an average server or VPS? Or are you going to implement something similar to a trading policy like you did in your previous RDF article?
No, I just choose the best one, or best 5
Yes, it can be discussed endlessly. Until the results improve - I do. And everything works fast
I'm not saying that this is the best approach, just need to check
we mean that a non-linear classifier with a kernel will work better. You can add layers to the neural network, but the feature space will be straggly separable and it will be retrained
you can add some KPCA or line disc. analysis for filtering, but then the process will again become uncontrollable and dependent only on these algorithms
mb, I'm missing something, but the core + fast nonlinear models it should be very coolSo have you already completed it or yet to add something more before publishing?
If it is, it is fast and in the process of learning and execution, then probably the only task to complete is to make the algo converge automatically over time and make the convergence faster.
So have you already completed it or yet to add something more before publishing?
If it is, it is fast and in the process of learning and execution, then probably the only task to complete is to make the algo converge automatically over time and make the convergence faster.
not ready yet, I just learning different kernels
not ready yet, I just learning different kernels
I am also trying to learn about the kernel trick and it's implementation, advantages and feasibility of application in modern servers and VPS using MQL5.
I just had a overall look into the above article about OPenCl. From your understanding so far about kernel trick can you please answer the 2 queries:
1.Is it mandatory to have GPU in a VPS to use this kernel trick method and it's relevant MQL5 systems or it can work with normal VPS with CPU?
2.Since it is trying to optimise the system using kernel memory, will it actually work in a regular VPS or any settings adjustment need to be done inside the registry files of the VPS before using this system?
I am also trying to learn about the kernel trick and it's implementation, advantages and feasibility of application in modern servers and VPS using MQL5.
I just had a overall look into the above article about OPenCl. From your understanding so far about kernel trick can you please answer the 2 queries:
1.Is it mandatory to have GPU in a VPS to use this kernel trick method and it's relevant MQL5 systems or it can work with normal VPS with CPU?
2.Since it is trying to optimise the system using kernel memory, will it actually work in a regular VPS or any settings adjustment need to be done inside the registry files of the VPS before using this system?
Its auto link from forum, not from me ^) sure I meant kernel tricks, not gpu kernels
Its auto link from forum, not from me ^) shure I meant kernel tricks, not gpu kernels
Ok, got it.....
Thank you for the clarification. Though I am not new to MQL, but I am new to this forum features and usage and hence, sorry for the confusion.
So I assume your system will work in any VPS which I had a doubt and I will also put more time to study kernel trick implementation in MQL5 which might help me in future to edit the source code if required after you publish the article.
Ok, got it.....
Thank you for the clarification. Though I am not new to MQL, but I am new to this forum features and usage and hence, sorry for the confusion.
So I assume your system will work in any VPS which I had a doubt and I will also put more time to study kernel trick implementation in MQL5 which might help me in future to edit the source code if required after you publish the article.
Good tutorial about ktricks https://pythonprogramming.net/kernels-with-svm-machine-learning-tutorial/
29,30,31 videos
sure you can use vps
Good tutorial https://pythonprogramming.net/kernels-with-svm-machine-learning-tutorial/
Thanks for the link..
I will keep learning and see the possibilities of implementation of kernel trick in MQL5.
Ну, дык там об этом и речь. Любой интерполяционный многочлен непригоден для экстраполяции. Фурье в точности повторяет исходный ряд, а полиномы вроде Лагранжа или Тейлора порождают кривые с лавинообразным ростом скорости изменения цены. Сглаживание смягчает картину, но не сильно, да и неправильно это - теряется связь с первоисточником.
Есть простой, понятный и действенный метод экстраполяции, с интерполяцией никак не связанный. Тренд.