Skip to content

GnosticNeuron — MAGNET‑Specific Gnostic Algebra Neuron

Implements a deep‑learning neuron grounded in gnostic algebra with characteristic activation transformations and parameter adaptations.


Overview

GnosticNeuron is a MAGNET‑specific neural network component that operates on the gnostic algebra framework. It introduces a centered coordinate system with parameters s (scale) and z0 (center), and supports both forward and backward gnostic transformations through activation functions fi, fj, hi, hj, and ei. The neuron maintains the gnostic characteristics:

  • fi = sech(2θ) (fidelity output)
  • fj = cosh(2θ) (infidelity entropy)
  • hi = tanh(2θ) (relevance output)
  • hj = sinh(2θ) (irrelevance output)
  • ei = 1 - fi (gnostic entropy)

Centered coordinate: θ = (x - z0) / S, where S is the scale parameter and z0 is the center.

This neuron integrates seamlessly into MAGNET models, providing the core gnostic transformations for advanced deep‑learning applications.


Key Features

  • Gnostic Algebra Core: Forward/backward passes using fi, fj, hi, hj, and ei transformations
  • Centered Coordinate System: θ = (x - z0) / S with s (scale) and z0 (center) parameters
  • Dual‑Entropy Types: Configurable entropy output ('i' → [0,1), 'j' → [0,∞)) via ei parameter
  • Adaptive Optimizers: Integrated handling of s and z0 parameters by optimizers with specialized scaling
  • Flexible Activation Types: Works with both standard activations and gnostic versions
  • Parameter Management: Dedicated handling of s, z0, *_s, and *_z0 parameters through MAGNET optimizers

Parameters

Parameter Type Default Description
activation str | callable 'relu' Activation function name or callable; supports MAGNET‑specific gnostic activations
units int 1 Number of output units/neurons
s float 1.0 Scale parameter controlling coordinate spread S (used in θ = (x - z0) / S)
z0 float 0.0 Center parameter defining the origin of the centered coordinate system
use_gnostic bool True Enable MAGNET‑specific gnostic transformations if True

Methods

__init__(activation='relu', units=1, s=1.0, z0=0.0, use_gnostic=True)

Initialize a gnostic neuron with the specified activation, units, and gnostic parameters.

__call__(X, training=True)

Perform forward pass with gnostic transformations. When use_gnostic is True, applies the full gnostic algebra framework with fi, fj, hi, hj, and ei transformations using centered coordinates θ = (x - z0) / s. Otherwise, behaves like a standard neuron with the specified activation.

compile(loss=None, optimizer=None)

Configure the neuron (or model containing the neuron) with a loss function and optimizer, ensuring proper handling of s and z0 parameters.

fit(X, y, epochs=1, batch_size=32, verbose=True, callbacks=None)

Train the neuron on (X, y) data using the configured loss and optimizer, with specialized parameter scaling for s and z0 during updates.

predict(X)

Generate forward predictions for input X without any training updates, applying gnostic transformations if enabled.

evaluate(X, y)

Compute loss and metric values on (X, y) data, returning both for validation while preserving gnostic parameter adaptations.

get_weights()

Collect all trainable parameters from the neuron, including s, z0, and activation weights.

set_weights(weights)

Restore neuron weights from a flat list, ensuring proper ordering of s, z0, and activation parameters.


Example Usage

from machinegnostics.magnet import GnosticNeuron, MSE, SGD

# Create a gnostic neuron with MAGNET parameters
neuron = GnosticNeuron(
    activation='relu',
    units=1,
    s=1.5,
    z0=0.2,
    use_gnostic=True
)

# Compile neuron (part of a model)
neuron.compile(
    loss=MSE(),
    optimizer=SGD()
)

# Train the neuron
neuron.fit(X_train, y_train, epochs=10, batch_size=32)

# Evaluate performance
loss, metric = neuron.evaluate(X_val, y_val)
print(f"Validation Loss: {loss:.4f}")

# Generate predictions
preds = neuron.predict(X_test)

Author: Nirmal Parmar