Wavelakes

Large Signal Models (LSMs), DeepDecison's Wavelake technology, are similar to Large Language Models (LLMs) in its ability to ingest extreme-scale datasets and then respond in real-time. Recent commercial examples of LLM technology include ChatGPT-4 from OpenAI and Gemini from Google. 

Wavelakes are designed for machine-to-machine messaging rather than for chatbot purposes to simulate human conversation.

Wavelakes convert non-linear signal impulses into linear algebra which can be used for financial engineering use cases to include risk management, arbitrage trading, and behavioral finance.

Wavelakes may be integrated into legacy environments to process logs into moving averages or integrated into next generation LLM-LSM environments. It enables the creation of time-dependent real-time models by continuously parsing an organization’s logs and processing real-time structured messages. By enabling continuous event updates, a dynamic human-like understanding of these logs is achieved. The Wavelakes approach is built on deep learning methodology akin to LLM’s leveraging deep learning for natural language processing (NLP) purposes. 

• 1,000:1 to 1,000,000 compression ratios with the property of limiting compression file sizes
• Enables continuous event updates
• Enables parallelized operations
• Enables out of time processing 

• Complex non-linear networks can be moved to linear matrix operations as the event impulse is converted to a wavelet of linear values
• Linear matrix operations enable human-like understanding of internal network operations
• Enables deep multilayer neural networks to be easily collapsed to a few layers
• Enables operations not/xor/and to be performed on a single connection