Machine learning in trading: theory, models, practice and algo-trading - page 544

 
Maxim Dmitrievsky:

There is a possibility that your monitor confuses colors, sometimes it is difficult to distinguish because of the wrong calibration

What did you build? If it's not a secret

There's no such thing.

Remember how I told you that the pound is in the wrong color on the arbitrage chart

What did you build - it's a secret.

 
Renat Akhtyamov:

There is no such an option.

Remember how I told you that the pound on the arbitrage chart is the wrong color

what i have built - it's a secret


All my colors matched.

 

Not about the market, but very useful and about the general approach to effectively build a model (at any stage something can go wrong, and we do not even realize it):


 

Teacherless learning (clustering) and RL(reinforcement learning). In an attempt to reduce the parameters to be optimized. Has anyone thought about how clustering can be applied? The second one is more complicated, you need specialized packages

example: https://robotwealth.com/unsupervised-candlestick-classification-for-fun-and-profit-part-1/

https://robotwealth.com/unsupervised-candlestick-classification-for-fun-and-profit-part-2/

Unsupervised candlestick classification for fun and profit – part 1
Unsupervised candlestick classification for fun and profit – part 1
  • 2015.11.10
  • Kris Longmore
  • robotwealth.com
Candlestick patterns were used to trade the rice market in Japan back in the 1800’s. Steve Nison popularised the idea in the western world and claims that the technique, which is based on the premise that the appearance of certain patterns portend the future direction of the market, is applicable to modern financial markets. Today, he has a...
 
Maxim Dmitrievsky:

Teacherless learning (clustering) and RL (reinforcement learning). In an attempt to reduce the parameters to be optimized. Has anyone thought about how clustering can be applied? The second one is more complicated, you need specialized packages

example: https://robotwealth.com/unsupervised-candlestick-classification-for-fun-and-profit-part-1/

https://robotwealth.com/unsupervised-candlestick-classification-for-fun-and-profit-part-2/

I've started to think about reinforcement learning. It seems to me that this is what is needed for exchanges.
 
Aleksey Terentev:
I began to think about reinforcement learning. I think this is what is needed for the exchanges.

I, too, am mastering Python at the same time. R pisses me off. There was an ancient article by o_o about Kohonen's layer, he wrote something there on plusses and so on, without examples and development

https://www.mql5.com/ru/articles/1562

Рецепты нейросетей
Рецепты нейросетей
  • 2009.02.12
  • o_o
  • www.mql5.com
Не так давно - на заре технического анализа, когда компьютеры были далеко не у каждого биржевика - появлялись люди, которые пытались предсказывать будущие цены по ими же придуманным законам и формулам. Таких людей часто называли шарлатанами. Время шло, усложнялись методы обработки информации, и теперь очень сложно найти равнодушного к...
 
Maxim Dmitrievsky:

I, too, am mastering Python at the same time. R pisses me off. There was an old article by o_o about Kohonen's layer, he wrote something on plusses and so on, without any examples or development.

https://www.mql5.com/ru/articles/1562

I wrote on pluses qt and opennn, frankly speaking nothing further than mlp (opennn) has been developed there yet.
In general I can add you as a contributor to my repository, I will explain scheme how it works for me.
 
Aleksey Terentev:
I wrote on qt and opennn, to be honest nothing has been developed longer than mlp (opennn).
In general I can add you as a counterpart to my repository, I will explain scheme how it works for me.

I'm afraid I need to learn a little more first, I'm not that much of a programmer :) maybe later in the next year

 
Maxim Dmitrievsky:

Teacherless learning (clustering) and RL (reinforcement learning). In an attempt to reduce the parameters to be optimized. Has anyone thought about how clustering can be applied? The second one is more complicated, you need specialized packages

example: https://robotwealth.com/unsupervised-candlestick-classification-for-fun-and-profit-part-1/

https://robotwealth.com/unsupervised-candlestick-classification-for-fun-and-profit-part-2/

Clustering is also an interesting method. I think it should be applied before training of the model, because in this way you will eliminate parameters that are not correlated at all.
I also keep it in mind. But I can't find the time to do it.
 
Aleksey Terentev:
Clustering is also interesting method. I think it should be used before training of model, because in this way you can sift out parameters that are not correlated at all.
I also keep it in mind. I just can't get my hands on it.

especially if we use multidimensional clustering, we can try to feed vectors with attributes and vectors with lag, say, increments... so that we could divide them into groups - which attribute properties would correspond to which increments in the future

and then use this set for training of NS, for example... i.e. like datamining

yes, exactly before training... or as a separate thing for TS