Independent AI research
Intelligence,
recursively
reimagined.
A new neural architecture built on finite hierarchical self-similarity.
FRACTAL DEPTH05FINITE / RECURSIVE
SCROLL TO DESCEND ↓
01 / PREMISE
Complex systems become stable through self-similarity. What if neural networks did the same?
02 / ARCHITECTUREOne pattern.
One pattern.
Many scales.
Each level repeats a shared structural principle at a finer resolution—creating an inherent hierarchy instead of adding external complexity.
01
Shared form
02Hierarchical refinement
03Finite recursion
03 / THE SHIFT
Beyond brute-force scale.
→
04 / RESEARCH FRONTIERQuestions worth
Questions worth
going deeper for.
FractalPI is an active research direction. Our hypotheses are designed to be tested, measured and challenged.
Can stable structure enable more aggressive training?
TRAINING DYNAMICS
Can self-similarity act as inherent regularization?
GENERALIZATION
Can hierarchy improve parameter and memory efficiency?
EFFICIENCY
How does fractal depth change scaling behavior?
SCALABILITY
05 / RESEARCH LINEAGEScience before
Science before
the system.
FractalPI grows from years of research at the intersection of nonlinear dynamics, complex systems and intelligent computation.
FULL GOOGLE SCHOLAR PROFILE ↗2021
PROCEDia COMPUTER SCIENCE 192 · KES2021
PDF ↗
Face emotional responses correlate with chaotic dynamics of eye movements
Albert Śledzianowski, Krzysztof Urbanowicz, Wojciech Glac, Renata Slota, Maria Wojtowicz, Monika Nowak & Andrzej Przybyszewski
NONLINEAR DYNAMICSEYE TRACKINGEMOTION AI
2022
INTELLIGENT INFORMATION & DATABASE SYSTEMS
PDF ↗
Detecting true and declarative facial emotions by changes in nonlinear dynamics of eye movements
Albert Śledzianowski, Jerzy P. Nowacki, Andrzej W. Przybyszewski & Krzysztof Urbanowicz
.75ACCURACY.90ROC–AUC.80PRECISION
THE NEXT SCALE OF INTELLIGENCE
Go deeper.
Explore a new geometry for machine intelligence.