Skip to main content

AI models with adaptive memory management and strategic forgetting, inspired by Greek mythology and neuroscience.

Project description

theRiverLethe

AI models with adaptive memory management and strategic forgetting, inspired by Greek mythology and neuroscience.

Academic History of Titans Architecture

  • Background: Prior to Titans Architecture, significant research was conducted related to test-time model adaptation and fast-weighted methods, primarily building on two key concepts:

    Self-supervised Learning and Meta Learning formed the foundation for these approaches.

  • 2017-03-09 Model-Agnostic Meta-Learning (MAML): Pioneered fast adaptation to new tasks with minimal training examples, establishing core principles for adaptive models.
  • 2019-09-29 Test-time Training: Advanced the field by introducing methods for model adaptation during inference time without requiring complete retraining.
  • 2024-07-05 TTT-Linear/MLP: Introduced as a Self-Attention alternative for Transformers, enabling automatic retention of input sequences by using Self-supervised learning.
  • 2024-12-31 Titans Architecture: Established a novel memory-based architecture combining Transformer's short-term memory capabilities (Self-Attention) with MLP-based long-term memory, significantly improving performance on extended sequential tasks.
  • 2025-MM-DD Atlas Model

Overview

Titans Origin

  • Titans introduced a new architecture family which consisted of:
    • Core Self-Attention (Short-term Memory, In-context learning)
    • Contextual Memory (Long-term Memory)
    • Persistent Memory (Fixed Memory)

Memory as a Context (MAC)

Memory as a Gate (MAG)

Memory as a Layer (MAL)

Atlas Model

  • Atlas is a Titan in Greek mythology who is assigned the role of supporting the celestial sphere.
  • Atlas Model proposes a combination of cognitive-scientific memory components.
    • Semantic Memory (Retrospective, Long-term Memory)
    • Episodic Memory (Retrospective, Long-term Memory)
    • Intentional Memory (Prospective, Long-term Memory)
    • Active Cognitive State (Working/Operational, Short-term Memory)

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

thelethe-0.0.1.tar.gz (6.3 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

thelethe-0.0.1-py3-none-any.whl (12.9 kB view details)

Uploaded Python 3

File details

Details for the file thelethe-0.0.1.tar.gz.

File metadata

  • Download URL: thelethe-0.0.1.tar.gz
  • Upload date:
  • Size: 6.3 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: python-requests/2.32.3

File hashes

Hashes for thelethe-0.0.1.tar.gz
Algorithm Hash digest
SHA256 392bdf45f89a3e35fe4dc584b98333f30bbd7bb7982ed8f87da9b640597c740e
MD5 6e2229451af9a05fcc6355cb1893e665
BLAKE2b-256 3a2a2a60738e7784024b03c6702bf12038afda85f243f0edd61817facb87505b

See more details on using hashes here.

File details

Details for the file thelethe-0.0.1-py3-none-any.whl.

File metadata

  • Download URL: thelethe-0.0.1-py3-none-any.whl
  • Upload date:
  • Size: 12.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: python-requests/2.32.3

File hashes

Hashes for thelethe-0.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 593fba8d3c3cb2da8e03ca59e18dc4cb9398d054d7f554c0efd5a272355085cc
MD5 268b0a7b5e6ef2ba162a150aff3fed76
BLAKE2b-256 7762bddbbd4371e6a50933c862de2498f4406e91e97c3cd143c0c684bdf213ae

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page