Privacy-Preserving Active Learning for sustainable aquaculture monitoring systems for extreme data sparsity scenarios

Description and use

A pipeline combining active learning with diversity regularization, DP‑SGD, and agentic orchestration reduces required labels by ~35% while maintaining accuracy and providing differential privacy in extreme sparsity, edge‑only aquaculture monitoring.

Problem

Only a handful of labeled sensor readings were available from 40 deployed ESP32 water‑quality sensors, while the rest were miscalibrated, destroyed, or unlabeled, creating extreme data sparsity and privacy‑sensitive telemetry.

Outcome

The pipeline reduced required labels by ~35% while maintaining F1 performance and providing formal differential privacy guarantees in an extreme sparsity, edge‑only setting.

Workflow

• Deploy ESP32 sensors with local 1D convolutional encoder to generate 32‑dim embeddings
• Use uncertainty sampling with diversity regularization to select unlabeled points for labeling
• Apply DP‑SGD with calibrated Gaussian noise during training
• Federate model updates from edge devices while accounting for privacy budget
• Run an agentic scheduler that monitors drift, budget, and decides next actions

Evidence

Out of 40 sensors deployed across six ponds, only 11 returned labeled, trustworthy readings after the first month.

Source

With lambda_div=0.5, the model reached the same F1 score with roughly 35% fewer labels than pure entropy sampling.

Source

DP‑SGD: clip per‑sample gradients, then add Gaussian noise calibrated to the clipping norm.

Source