Cadence

Website · Documentation · Application demos · Paper · PyPI

An experimental brain that learns, remembers, imagines and acts.

Cadence aims to build a simulated human-like brain from simplified biological mechanisms. Its default System 1 is a continuing animal-like brain with perception, plastic connections, memory and action. Optional System 2 adds recursive feedback through observing cortical regions in the same neural graph. The base can already be deep and modular.

Bounded, observer-like regions carry local state, communicate through ports, read back activity and retain records. They repair disagreement until the whole brain settles into one coherent state; actual observations and consequences guide learning. A Cadence brain can be deep. It learns without backpropagation, through local contrasts between free and nudged activity. Actions require a qualified equilibrium; teaching has its own finite or qualified phase contract. A settled answer can still be wrong about the world, so capability is measured through free behavior. Cadence is alpha research software, not a claim of human-level intelligence.

These pages describe Cadence 0.71.0, including opt-in qualified teaching, calibration and work reports. Install the published package for the examples below, or use the source installation.

How a Cadence brain differs from a feed-forward network

A feed-forward deep network computes an answer in one pass. Activity moves from the input layer to the output layer, every layer is evaluated once, and nothing travels back while the answer forms. Training uses backpropagation: an error measured at the output is sent backwards through all layers as a chain of derivatives, and every weight moves by its share of that single global error.

A Cadence brain reaches its answer by settling. Its regions are connected in both directions, and every neuron keeps moving its own state toward what its inputs and its neighbours drive it to. This local repair repeats until every neuron agrees with its neighbours within a tolerance. The answer is that equilibrium of the whole brain: a consensus among local patches, reached through local repair alone. A later region shapes an earlier one while the answer forms, and a brain that does not settle refuses to act.

Learning uses the same local dynamics. A nudge at the motor neurons points toward a demonstrated answer or along the action just taken. Each synapse compares the activity of its own two neurons between free and nudged phases; the label or measured outcome sets the update. The default teaching and reward eligibility phases run for a finite duration. Set LearnerConfig(qualified=True, ...) to require the full free and teaching equations before accepting a supervised update. A refused lesson keeps parameters and optimizer history intact. When the situation returns, the brain settles with its changed connections. No error is sent backwards through a stack of layers, and the brain keeps no backward computation graph.

Cadence brains can be deep. Brain.compose(inputs=4, actions=2, modules=(64, 32, 16)) chains three processing regions, each exchanging activity with the next, and all of them settle together. Depth adds regions to the one settlement. The learning rule stays local at every depth.

Why Cadence

Animal and human brains learn from experience and not by backpropagation with gradient descent. Cadence is designed the same way. Synapses change from the activity of the two neurons they connect, reward scales that change, and memory writes are local too. The brain has no separate training mode.

For embodied AI this design gives:

• Learning on the job. The same brain acts and learns from every measured outcome. There is no difference between training and inference.
• Adaptation to changed conditions. A changed body or world shows up in what the brain measures, and the live brain adjusts without being told what changed.
• Local learning. No backward pass through the network is needed, so learning can run where the brain runs.
• Memory. A working trace carries the recent past, and fast and persistent associative memory keep what mattered.
• Settled answers. Every action is a qualified settled state of the whole brain. A brain that does not settle refuses to act.
• Inspection. Region activity can be read while the brain runs, and private imagination tests a response before the body commits to it.

Under idealized conditions the local contrast follows the gradient that backpropagation would compute. The learning rule states those conditions, and the comparison is of update mechanisms. These properties are shown in simulation in the examples. An advantage on a physical robot is a separate test.

Start with System 1

Python 3.11+ and NumPy are required. Install the published release for this basic example:

python -m pip install cadence-net==0.71.0

Its released documentation describes the APIs included in that package.

import numpy as np
from cadence import Brain

brain = Brain.compose(inputs=4, actions=2, modules=(16, 8), seed=7)
observation = np.array([[1.0, 0.0, 0.0, 0.0]])
action = brain.step(observation)

# A tiny environment rewards action 0 and supplies the next observation.
reward = (action == 0).astype(float)
next_observation = np.array([[0.0, 1.0, 0.0, 0.0]])
action = brain.step(next_observation, reward=reward, done=np.array([False]))
assert action.shape == (1,)

step learns from the preceding action's measured reward, then chooses the next action. teacher= can label the current observation. Keep each batch row attached to the same life. There is no training/inference mode switch. Continuous interaction covers teaching, resets and saved continuation. The continuing brain example combines actual rewards, current teacher labels and a checkpoint awaiting feedback on 0.71.0.

The constructor includes a working trace and fast/persistent associative memory. The trace carries recent activity; learned graph parameters and consolidated associations retain changes across resets. A current teacher changes graph parameters. Actual chosen-action outcomes write associative memory. act reads both memory pathways; independent predict and accuracy read neither, so they measure the graph's learned response. Capacity is finite; overlapping associations and further plasticity can interfere with recall.

phases = brain.imagine([observation, next_observation])
assert phases  # Inspect phase.converged before using an imagined response.

Imagination carries a private trace without changing live memory, random state or pending feedback. It evaluates responses to the observations you supply. For learned environmental consequences and action planning, use the separate temporal model.

To work from source, install from the library repository root:

python -m pip install -e .

This installs the local source as an editable package.

Add optional System 2

recursive = Brain.compose(
    inputs=4, actions=2, modules=(16, 8), observers=(8,), seed=7,
)

Observer regions read and return influence to the base, motor regions and earlier observers. They join the same settlement and use the same interaction interface. This makes recursive feedback available; learning when it helps remains a task for experience and evaluation.

Actions and independent predictions require the full neural equation residual to meet the configured tolerance. Exhausting the budget refuses an action without changing its live state, memory or pending feedback. If step has learned a real outcome before the next action refuses, retry act without submitting that reward again. Numerical damping stays within the total budget and checks the original equations. It does not change the teaching rule. See contracts.

Go further

Build a brain for custom wiring, memory for traces and associations, and the memory/planning example for a bounded demonstration with actual toy-body outcomes. Record patches provide event records and consolidation. The advanced population solver provides exact state-and-error readback under its own numerical contract.

cadence-demos contains the active application demos. cadence-examples preserves research examples and viewer tools with their own declared library pins.

Documentation · API · Contributing · Changelog · Research tasks

Licensed under GPL-3.0.