Synapse - Create Config

PRESET TEMPLATES

quick-start: pre-fills all fields for a standard problem

CONFIG NAME

Name (no .json):

TOPOLOGY

TRAINING

Should Train

Activation function:

WEIGHTS

Load Weights Save Weights

Weight range: low high

FILES

Save weights to:

Load weights from:

Truth table:

Test inputs:

Inputs:

LEARNING

Learning rate:

how fast weights adjust (Adam optimizer, try 0.001 to 0.01)

Max iterations:

hard cap on training passes; training also stops early on error threshold

Error threshold:

training stops when average output error falls below this value

TESTING

auto-filled from architecture -- override as needed

Test cases: Outputs:

Test input rows: Cols:

Input rows: Cols:

DISPLAY

Truth Inputs Test Inputs

Steps Memory Weights

Keep alive (ms):

print training progress every N iterations (0 = silent, 5000 = every 5k steps)

> json preview