Great news, everyone! We released Apheris 3.3. With prior releases, we set the foundation for two really useful improvements for our users today — derived Datasets and Multi-Gateway Organizations. Yet, because our team is a dedicated bunch of awesomeness, we also implemented a new model and many improvements to streamline the data science experience further. Ready? Let’s dive in…
TL;DR
Derived Datasets: Allow you to persist pre-processing results safe and sound on a Compute Gateway, speeding up your workflows by eliminating the need to repeat pre-processing steps;
Multi-Gateway Organizations: Manage multiple Compute Gateway deployments within a single organization. Allow governed workloads across your data estate and underlying infrastructure;
Federated Logistic Regression: Addition to the Regression model family in the Apheris Model Registry.
Apheris Statistics upgrade: Added additional Python SDK functionality including PCA Transformation and Covariance Metrix generation.
In Detail
We grow with our customer needs and boy did they grow. Many of our customers run or plan to run quite complex analytical and machine learning workloads in a federated manner. When data is distributed across organizational boundaries (and sometimes also within the same org), it is likely that these datasets do not have the same structure or require other pre-processing steps.
The new Derived Datasets feature lets you persist intermediate pre-processing results on the same Compute Gateway. This new feature eliminates the need to repeat pre-processing steps and accelerates the speed at which you can develop your models.
Derived datasets are governed automatically by the same Asset Policy as the original dataset. Of course, a Derived Dataset can only be created if approved by the Data Custodian and, as usual, all steps are logged by the Compute Gateway for traceability and compliance.
Multi-Gateway Organizations enable the data custodian side to manage multiple Gateways within the same logical Organization. A Compute Gateway lives close to where sensitive data resides but this might mean data lives in different environments or infrastructure (on prem, various cloud providers, etc.). Prior to v3.3, Apheris offered a 1:1 relationship between Gateway and Organization but with the growing projects of our customers, this had to change. To simplify the management of such use cases, data custodians can now manage their whole data estate from one organization with a single source of truth.
Federated Logistic Regression got added to the regression model package in the Apheris Model Registry and is ready for federated action.
Part of the Apheris Regression Models repository
A generalized logistic regression model with build in federated feature selection using Principal Components Analysis
Training and validation released with Apheris 3.3.
Also the Statistics Package got an upgrade with two new functions: pca_transformation for federated Principal Components Analysis and cov for federated calculation of the Covariance matrix.
We will continue to add models to the Model Registry and, of course, you can always add your own models to your organization’s deployments. As Apheris uses NVIDIA FLARE under the hood, you can comparatively easily port models from commonly used model libraries such as Hugging Face or BioNeMo.
By the way, our team is hiring a Senior ML Engineer and a Senior ML Field Engineer. If you want to allow the world to safely collaborate on data, join us! Check out all open vacancies
As a famous chef once said, a great product is small things done well. Apheris 3.3 provides further improvement for the data science workflow and daily work with the Apheris product. If you are not benefitting from Apheris Compute Gateways already, get in touch and schedule a demo. We are happy to chat.
