π-Gen Project

introducing human-like learning to reshape AI

Inductive Learning (IL)

Learn by inducing general rules

• DonDe algorithm: induce general rules for precision decision-making in high dimensional space (v1.0 finished)
• FEVID algorithm: feature value importance detection

Self-Inductive Learning (SEIL)

Learn by inducing general rules by oneself, SEIL is IL plus KAL

• VIC-SEIL algorithm: generalisable and explainable image recognition

Knowledge-Augmented Learning (KAL)

Apply existing knowledge (like grammar) to drastically boost learning efficiency and leverage Logical Reasoning (LR) for reliable reasoning.

• GLOR algorithm: precise language comprehension, apply grammar to resolve scaling law and LR to eliminate hallucinations

IL&KAL&LR

IL&KAL are human-like learning and Logical Reasoning (LR) is reliable reasoning, they together will be the fourth learning method after Supervised Learning, Unsupervised Learning, Reinforcement Learning. IL&KAL&LR are generalisable, explainable, transferable, reliable, verifiable, adaptable to small datasets, and enable highly efficient learning for small models.

Name of π Gen

π Gen (π Genius, PAI Gen, Profound AI Genius).