Originally reported by devclass.com. Published September 2019.
Neural Structured Learning (NSL) for TensorFlow
The TensorFlow team introduced Neural Structured Learning (NSL), a new framework designed to let both novice and experienced developers train neural networks using structured signals alongside conventional feature inputs.
According to the team, NSL can be used to build accurate and robust models for a range of domains, including computer vision, natural language understanding, and general prediction tasks. The key differentiator is its use of structured signals, which can take two forms:
- Explicit structure: Represented by a graph, such as a knowledge graph or social network graph connected to training data.
- Implicit structure: Induced by adversarial perturbation, where the model is regularized against small, structured input changes to improve robustness.
By combining graph-based neighbor signals with standard feature inputs during training, NSL aims to improve model generalization, particularly in scenarios where labeled data is scarce but relational data is available.
TeamCity 2019.1.3: 50 Fixes Including Performance and Security
JetBrains released TeamCity 2019.1.3, a maintenance update for its continuous integration platform that addresses 50 reported issues. The release is notable for the breadth of fixes it delivers, spanning performance, memory management, and security.
Performance and Memory Fixes
Six of the fifty fixes were performance-related. Key improvements include:
- Reduced excessive memory consumption in the XML-RPC request processor.
- Fixed a memory leak caused by the
getNodesWithoutChildrenmethod. - Resolved two separate bugs that were degrading the speed of health report generation within the platform.
Security Fixes
Three security vulnerabilities were also patched in this release. JetBrains did not publicly disclose the details of these issues, which is standard practice until affected users have had time to upgrade.
AWS CloudWatch Container Insights Reaches General Availability
Amazon Web Services announced that CloudWatch Container Insights — its operational monitoring service for containers and containerized applications — had reached general availability (GA).
The service was first announced as an open preview in July 2019, initially covering new clusters running on Amazon ECS and AWS Fargate. The GA launch expanded the scope significantly:
- Support extended to existing clusters, not just newly created ones.
- Coverage added for Kubernetes and Amazon ECS for Kubernetes.
- Amazon ECS and AWS Fargate support retained from the preview.
CloudWatch Container Insights automatically discovers all running containers within a cluster and collects performance and operational metrics at every layer of the container stack. This gives operators a unified view of CPU, memory, disk, and network usage across their containerized workloads without requiring manual instrumentation.
GitLab 12.2.4 Patch Release
GitLab issued a patch release — version 12.2.4 — covering both its Community Edition (CE) and Enterprise Edition (EE). The release addressed bugs and regressions introduced in the 12.2 series.
What Was Fixed
- Nine bugs affecting both Community Edition and Enterprise Edition were resolved.
- An Enterprise Edition-specific bug was fixed where group hooks failed to fire in the
PostReceivehook handler.
Upgrade Path
GitLab confirmed that 12.2.4 does not require any new database migrations or scheduled downtime, making it a low-risk upgrade for teams running the 12.2.x series.