I Trained a Neural Network on Real Medical Data and Fit It in 8.9 kB of Pure C

Short summary

The author discovered Hasaki, their embedded neural network trainer, supports regression modeling. Testing with 3,658 real patient records from the Framingham Heart Study, they trained a model predicting systolic blood pressure with 7.8 mmHg mean absolute error. The quantized INT8 model exports as self-contained C code: 8.9 kB, zero runtime dependencies, ready for microcontroller deployment.

• •Hasaki embedded neural network trainer can do regression—an undocumented feature the author discovered through testing
• •Trained on 3,658 real Framingham patient records, achieving 7.8 mmHg mean absolute error predicting blood pressure
• •Final model: 8.9 kB pure C, INT8 quantized, zero runtime dependencies for direct microcontroller deployment

Generated with AI, which can make mistakes.