Navigating through complex photonic design space using machine learning methods
Bibliographic record
Abstract
The performance and functionality of integrated photonic devices can be enhanced by using complex structures controlled by a large number of design variables. However, the optimization of such high-dimensional structures is challenging, often limiting their realization. Global optimization algorithms and artificial neural networks are increasingly used to tackle these problems. Although these are exciting new developments, the outcome is a single optimized design meeting particular performance objectives selected upfront. The influences of the various design parameters remain hidden. Here we report on our strategy of using machine learning pattern recognition techniques to create a methodology for building the global performance map of a high-dimensional design space. As an example and demonstration, we study the design of a vertical grating coupler consisting of silicon and subwavelength metamaterial segments. We show how the relationship between designs with comparable primary performance can be clearly revealed by identifying the minimum number of characterizing parameters that defines the subspace of good designs, significantly scaling down the complexity of the problem. Moreover, the subspace can be identified using only a small number of good design solutions. We reveal design areas with comparable fiber coupling efficiency but with significant differences in other performance criteria, such as back-reflections, tolerance to fabrication uncertainty and minimum feature size. This novel approach provides the designer with a global perspective of the design space, enabling informed decisions based on the relative priorities of all relevant performance specifications and figures-of-merits for a particular application. Insights from the mapping exercise also inspired new design structures with enhanced characteristics.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".