Evaluación de prestaciones de las redes de cápsulas matriciales sobre distintos escenarios de clasificación de imagen
Bibliographic record
Abstract
[SPA] En este documento se redacta un estudio realizado sobre redes neuronales de cápsulas. Este tipo de redes fueron inventadas por un grupo de investigadores de la universidad de Toronto con Geoffrey E. Hinton a la cabeza y con la colaboración especial de Sara Sabour y Nicholas Frosst. Su invención en 2017 supuso una mejora importante en cuanto al avance de la investigación del machine learning. Cabe destacar que las redes de cápsulas se pueden dividir en dos grupos: las redes de cápsulas vectoriales y las matriciales, siendo las cápsulas matriciales una evolución de las otras y publicadas en 2018. Los siguientes apartados trataran en primer lugar, unos conceptos básicos de aprendizaje máquina y de las limitaciones de las redes neuronales anteriores, por las cuales la comunidad científica comenzó a investigar en el avance de nuevas tecnologías, hasta dar lugar a las redes de cápsulas. Posteriormente, se explicarán teóricamente los tipos de redes de cápsulas. Para finalizar, se realizará una comparación de estas redes de manera experimental. \n[ENG] In this document a study which has been carried out on capsule neural networks is written. These types of networks were invented by a group of researchers from Toronto University with Geoffrey E.Hinton at the head of it and with special collaboration from Sara Sabour y Nicholas Frosst. Their invention in 2017 was an important improvement regarding the advances in the research of the machine learning. It should be noted that capsule neural networks can be divided into two groups: vector capsules and matrix capsules, being matrix capsules an evolution of the other ones and published in 2018. The following sections are first of all about basic concepts of machine learning and limitations of previous neural networks, whereby the scientific community started to research in the advance in new technologies to result in capsule networks. Subsequently, the types of capsule networks will be explained theoretically. Finally, a comparison of these networks will be made experimentally.
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 | 0.014 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.006 | 0.002 |
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".