Evaluating Abstract Generalised Networks (AGNs)

AGNs function as unified systems that integrate predictions across multiple timeframes simultaneously. This holistic approach reveals how information propagates through different timeframes, uncovering interconnected dynamics that traditional single-horizon analysis misses.

AGNs have the ability to identify short-term fluctuations as well as long-term trends, providing a comprehensive view of system behaviour that enables more informed decision-making.

Data Requirements

AGNs perform optimally when provided with highly granular time series data. The effectiveness of its propagation methodology fundamentally depends on its ability to identify subtle, early changes in a system's dynamics and track how information evolves across different time scales.

AGNs are limited by low-frequency data. When data points are recorded infrequently, AGNs cannot generate the detailed insights needed to outperform traditional forecasting methods. For optimal results, the data used should have the highest practical recording frequency for your specific domain.

Core Evaluation Framework

Predictive Lead Time

Assess the AGNs' foresight by quantifying its lead time - how many minutes, hours, or days before a significant system event they provide consistent directional signals. This demonstrates its capacity to anticipate systemic shifts before they fully manifest.

Reaction Speed

Measure the AGNs' adaptability by evaluating how quickly it adjusts predictions following sudden system shocks. This is calculated as the time elapsed between a shock event and the point where the AGNs' outputs align with the new system state.

Time Horizon Alignment

Verify the internal consistency of AGNs by confirming that predictions for nearby time horizons (e.g. 1, 2 and 3 minute forecasts) are coherent and build upon each other logically, without contradictions or conflicting signals.