Discussing the article: "Neural Networks in Trading: An Agent with Layered Memory"

New comment
 

Check out the new article: Neural Networks in Trading: An Agent with Layered Memory.

Layered memory approaches that mimic human cognitive processes enable the processing of complex financial data and adaptation to new signals, thereby improving the effectiveness of investment decisions in dynamic markets.

The growing volume of financial data requires traders not only to process it rapidly but also to analyze it deeply in order to make accurate and timely decision. However, the limitations of human memory, attention, and the ability to handle large amounts of information can lead to missed critical events or erroneous conclusions. This creates a need for autonomous trading agents capable of efficiently integrating heterogeneous data - quickly and with high precision. One such solution has been proposed in the paper "FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design".

The proposed FinMem framework is an innovative large language model (LLM)-based agent that introduces a unique multi-level memory system. This approach enables efficient processing of data with varying types and temporal significance. The FinMem memory module is divided into a working memory, designed for short-term data processing, and a stratified long-term memory, where information is categorized according to its relevance and importance. For instance, daily news and short-term market fluctuations are analyzed at a superficial level, while reports and studies with long-term implications are stored in deeper memory layers. This structure allows the agent to prioritize information, focusing on the most relevant data.


The profiling module in FinMem allows the agent to adapt to professional contexts and market conditions. Taking into account individual preferences and the user's risk profile, the agent tailors its strategy for maximum efficiency. The decision-making module integrates current market data with stored memories to generate well-reasoned strategies. This enables the consideration of both short-term trends and long-term patterns. Such a cognitively inspired design allows FinMem to remember and utilize key market events, thereby increasing the accuracy and adaptability of its decisions.


Author: Dmitriy Gizlyk

 
Hello, interesting article. Unfortunately I can't compile the Research.mq5 file - the line if((!CreateDescriptions(actor, critic, critic))) - Incorrect number of parameters. I can't move further(
 
djgagarin #:
Hello, interesting article. Unfortunately I can't compile the Research.mq5 file - the line if((!CreateDescriptions(actor, critic, critic))) - Incorrect number of parameters. I can not move on(

Good afternoon, From which catalogue is the Research file loaded? There are indeed a lot of parameters. Only one model is used in this work.

 
Dmitriy Gizlyk #:

Good afternoon, From which catalogue is the Research file downloaded? There are indeed a lot of parameters here. In this paper only one model is used.

I have looked through the catalogues and I am already confused where I got it((
Can you please direct me which catalogue to use for this paper?

 
djgagarin #:

On catalogues lazily and already confused where I took((
Please direct me what catalogue to use for this article?

All files related to this article are located in the FinMem folder.

 

Tried all sorts of things but didn't come up with your results.

I'm sorry, can you give proper instructions on what to run and what files in what order.

Thank you.

New comment