The Ontario Data Safe Haven: Bringing High Performance Computing to Population-wide Data Assets
Bibliographic record
Abstract
IntroductionCanadian provincial health systems have a data advantage – longitudinal population-wide data for publicly funded health services, in many cases going back 20 years or more. With the addition of high performance computing (HPC), these data can serve as the foundation for leading-edge research using machine learning and artificial intelligence. Objectives and ApproachThe Institute for Clinical Evaluative Sciences (ICES) and HPC4Health are creating the Ontario Data Safe Haven (ODSH) – a secure HPC cloud located within the HPC4Health physical environment at the Hospital for Sick Children in Toronto. The ODSH will allow research teams to post, access and analyze individual datasets over which they have authority, and enable linkage to Ontario administrative and other data. To start, the ODSH is focused on creating a private cloud meeting ICES’ legislated privacy and security requirements to support HPC-intensive analyses of ICES data. The first ODSH projects are partnerships between ICES scientists and machine learning. ResultsAs of March 2018, the technological build of the ODSH was tested and completed and the privacy and security policy framework and documentation were completed. We will present the structure of the ODSH, including the architectural choices made when designing the environment, and planned functionality in the future. We will describe the experience to-date for the very first analysis done using the ODSH: the automatic mining of clinical terminology in primary care electronic medical records using deep neural networks. We will also present the plans for a high-cost user Risk Dashboard program of research, co-designed by ICES scientists and health faculty from the Vector Institute for artificial intelligence, that will make use of the ODSH beginning May 2018. Conclusion/ImplicationsThrough a partnership of ICES, HPC4Health and the Vector Institute, a secure private cloud ODSH has been created as is starting to be used in leading edge machine learning research studies that make use of Ontario’s population-wide data assets.
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 | 0.012 | 0.046 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.008 |
| Science and technology studies | 0.006 | 0.005 |
| Scholarly communication | 0.012 | 0.013 |
| Open science | 0.006 | 0.018 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.025 | 0.008 |
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".