Skip to content

Layers (Machine Gnostics MAGNET)

The machinegnostics.magnet module provides layer implementations for constructing MAGNET networks.


Key Layer Classes

  • Dense — Fully connected affine layer (y = x @ W + b)
  • iDense — Gnostic estimating weight dense layer (S, lel)
  • jDense — Gnostic quantifying weight dense layer (S, lel)
  • BatchNorm — Standard batch normalization (gamma, beta, running stats)
  • GnosticBatchNorm — Batch norm with gnostic weighting (kind="i"/"j")
  • Flatten — Flatten structured inputs to (batch, -1)
  • Layer — Abstract base for MAGNET layers (params, grads, parameters())

Author: Nirmal Parmar