Error-Modulated Hebbian Updates

• Error-modulated Hebbian updates are synaptic plasticity rules that factorize learning into local presynaptic-postsynaptic correlations modulated by external error or reward signals.
• They derive modulators from various objectives—such as contrastive goodness, kernel similarity, and stability measures—allowing local, task-sensitive weight adjustments.
• Applications in deep networks, spiking systems, and unsupervised representation learning show improvements in performance and stability across diverse learning scenarios.

Error-modulated Hebbian updates are synaptic plasticity rules in which a local Hebbian correlation is multiplied by a third factor carrying error, reward, teaching, or stability information. In canonical form, a synapse $i \to j$ obeys $\Delta w_{ij}=\eta\,m\,y_j x_i$, or $\Delta w_{ij}=\eta\,m\,g_j x_i$ when the postsynaptic term includes a local derivative or surrogate factor. This three-factor structure appears across contemporary local-learning research: in Forward-Forward learning with squared-activation goodness, in policy-gradient plasticity for spiking neurons, in layer-wise Information Bottleneck training with working-memory-dependent modulators, in quantum probability-flow objectives, and in microcircuit models where dis-inhibitory inhibition carries the effective error signal (Terres-Escudero et al., 2024, Bartlett et al., 2019, Daruwalla et al., 2021, Ohzeki, 1 Jun 2026, Rossbroich et al., 2023).

1. Canonical three-factor structure

The defining property of an error-modulated Hebbian rule is the factorization of plasticity into a presynaptic term, a postsynaptic term, and a modulatory signal. In its simplest form,

$\Delta w_{ij}=\eta\,m\,y_j x_i,$

where $x_i$ is presynaptic activity, $y_j$ is postsynaptic activity, $m$ is a modulatory signal, and $\eta>0$ is a learning rate. A common variant replaces $y_j$ by a local postsynaptic drive $g_j$, such as $\Delta w_{ij}=\eta\,m\,y_j x_i$0, yielding

$\Delta w_{ij}=\eta\,m\,y_j x_i$1

The same structure is written in continuous time as

$\Delta w_{ij}=\eta\,m\,y_j x_i$2

with $\Delta w_{ij}=\eta\,m\,y_j x_i$3 as the third factor. In spiking formulations derived from policy gradients, the local Hebbian-like factor often becomes $\Delta w_{ij}=\eta\,m\,y_j x_i$4, where $\Delta w_{ij}=\eta\,m\,y_j x_i$5 is the realized spike, $\Delta w_{ij}=\eta\,m\,y_j x_i$6 is its stochastic firing probability, and the centered term $\Delta w_{ij}=\eta\,m\,y_j x_i$7 plays the role of a postsynaptic surprise variable (Terres-Escudero et al., 2024, Bartlett et al., 2019).

This structure differs from both pure Hebbian learning and backpropagation. Pure Hebbian plasticity uses only two factors,

$\Delta w_{ij}=\eta\,m\,y_j x_i$8

and therefore lacks an explicit task-dependent gate. Backpropagation, by contrast, computes updates from gradients of a global loss propagated through a reverse pathway. Error-modulated Hebbian rules occupy the intermediate regime: they preserve synapse-local update algebra, but the direction and magnitude of plasticity are controlled by an additional signal encoding reward, error, target mismatch, stability margin, or validation.

2. Objective-derived modulators

In many recent formulations, the third factor is not introduced heuristically; it is derived from a local objective. In the Forward-Forward Algorithm (FFA), a layer with pre-activation $\Delta w_{ij}=\eta\,m\,y_j x_i$9, activation $\Delta w_{ij}=\eta\,m\,g_j x_i$0, and goodness

$\Delta w_{ij}=\eta\,m\,g_j x_i$1

has local goodness gradient

$\Delta w_{ij}=\eta\,m\,g_j x_i$2

Positive examples increase goodness and negative examples decrease it, so the local update becomes

$\Delta w_{ij}=\eta\,m\,g_j x_i$3

where $\Delta w_{ij}=\eta\,m\,g_j x_i$4 is generated by the goodness-to-probability mapping. The paper gives two explicit mappings,

$\Delta w_{ij}=\eta\,m\,g_j x_i$5

and the corresponding derivative factors, such as $\Delta w_{ij}=\eta\,m\,g_j x_i$6 or $\Delta w_{ij}=\eta\,m\,g_j x_i$7, become explicit modulators multiplying the local correlation (Terres-Escudero et al., 2024).

A distinct layer-local derivation appears in HSIC-based Information Bottleneck learning. There the update decomposes as

$\Delta w_{ij}=\eta\,m\,g_j x_i$8

with local Hebbian term

$\Delta w_{ij}=\eta\,m\,g_j x_i$9

and modulatory term

$\Delta w_{ij}=\eta\,m\,y_j x_i,$0

Here the third factor is neither scalar reward nor explicit backpropagated error; it is a batch-statistic, layer-wise signal built from centered kernel similarities over an effective working-memory window (Daruwalla et al., 2021).

A third construction derives the modulator from a physics-based stability objective. In quantum probability-flow learning, one-spin-flip leakage channels define local gaps

$\Delta w_{ij}=\eta\,m\,y_j x_i,$1

and the measured survival loss produces instability weights

$\Delta w_{ij}=\eta\,m\,y_j x_i,$2

The resulting coupling update is

$\Delta w_{ij}=\eta\,m\,y_j x_i,$3

that is, Hebbian correlation $\Delta w_{ij}=\eta\,m\,y_j x_i,$4 gated by a softmax-like instability signal. In the high-temperature limit $\Delta w_{ij}=\eta\,m\,y_j x_i,$5, this reduces, up to a constant learning rate, to the classical Hebbian rule; in the low-temperature limit, learning concentrates on the weakest margin (Ohzeki, 1 Jun 2026).

Framework	Local Hebbian factor	Modulatory term
FFA	$\Delta w_{ij}=\eta\,m\,y_j x_i,$6 or $\Delta w_{ij}=\eta\,m\,y_j x_i,$7	Derived from $\Delta w_{ij}=\eta\,m\,y_j x_i,$8
HSIC Information Bottleneck	$\Delta w_{ij}=\eta\,m\,y_j x_i,$9	$x_i$0 from centered kernel statistics
Quantum probability flow	$x_i$1	$x_i$2 or real-time power-law gate

These examples show that the same algebraic form can arise from contrastive goodness maximization, kernelized compression objectives, or local stability margins. This suggests that the term “error” in this literature denotes a broader family of modulators than the conventional supervised residual.

3. Eligibility traces and temporal credit assignment

When rewards or supervised errors are delayed, error-modulated Hebbian rules are typically implemented through eligibility traces. In Bartlett and Baxter’s spiking reinforcement-learning formulation, the network maximizes the long-term average reward

$x_i$3

For a Bernoulli spiking neuron with $x_i$4, the local log-likelihood derivative is

$x_i$5

The synapse accumulates this in a leaky eligibility

$x_i$6

and plasticity is gated by the broadcast reward or reward-minus-baseline,

$x_i$7

The same paper shows that, under factorized joint policies and standard ergodicity and small-step-size assumptions, each neuron can update its own parameters using only local eligibility and the shared reward, yielding locally optimal performance without explicit inter-neuronal communication (Bartlett et al., 2019).

MOHQA uses a similar temporal mechanism, but with sparse STDP-inspired correlations. Its eligibility dynamics are

$x_i$8

where $x_i$9 depends on top-percentile or bottom-percentile pre-post correlations between DQN-derived features and a one-hot MOHN action head. The synaptic update is

$y_j$0

followed by clipping to $y_j$1. This architecture was designed for confounding POMDPs in which TD errors are inaccurate; the trace bridges temporal delays between salient feature–action events and sparse reward (Ladosz et al., 2019).

A more explicit gradient-free temporal rule appears in noise-based reward-modulated learning. There, noisy neurons induce a directional-derivative factor

$y_j$2

and the inter-reward eligibility becomes

$y_j$3

At reward time $y_j$4,

$y_j$5

with $y_j$6 as reward prediction error relative to a running average. The update remains synaptically local because it uses presynaptic activity, postsynaptic perturbation, and two global scalars, $y_j$7 and $y_j$8 (Fernández et al., 31 Mar 2025).

Online supervised spiking decoders extend the same logic to dense frame-wise errors. In the BCI setting, the instantaneous three-factor update is

$y_j$9

with dual-timescale traces

$m$0

$m$1

This formulation avoids backpropagation through time, uses $m$2 memory in sequence length, and couples rapid adaptation to slower consolidation (Nallani et al., 17 Sep 2025).

4. Synaptic specificity, circuit realization, and biological interpretation

A central biological issue is whether the third factor is merely global or whether it can be synapse-specific enough to preserve higher-order learning. The proofreading account of complex Hebbian learning argues that higher-order-correlation learning fails abruptly when synaptic updates are insufficiently connection-specific. In a one-unit ICA model with crosstalk matrix $m$3, the nonlinear Hebbian rule

$m$4

collapses above a modest threshold to a second-order, PCA-like solution. In a three-input example with one Laplacian and two Gaussian sources, the empirical threshold was $m$5 with $m$6, and across four mixing matrices the average was $m$7. To prevent this “error catastrophe,” the paper proposes a neocortical proofreading mechanism in which a validation factor gates plasticity: $m$8 or in STDP form,

$m$9

Here $\eta>0$0 is generated by a circuit involving thalamic relay cells, layer-4 spiny stellate cells, and layer-6 corticothalamic neurons, so that only independently corroborated pre-post coincidences are expressed as weight changes (Cox et al., 2010).

A different circuit-level solution embeds the error in inhibition itself. In the dis-inhibitory control framework, each local unit contains an excitatory neuron and an inhibitory interneuron with dynamics

$\eta>0$1

$\eta>0$2

At steady state, the local error decoded at the excitatory cell is

$\eta>0$3

and the exact learning rule becomes

$\eta>0$4

Its single-synapse form,

$\eta>0$5

is again a three-factor update. In the absence of inhibition it reduces to a postsynaptic-threshold rule,

$\eta>0$6

which the paper presents as consistent with in vitro phenomenological plasticity models (Rossbroich et al., 2023).

These two lines of work treat the third factor differently. Proofreading emphasizes selective approval that suppresses false updates caused by crosstalk, whereas dis-inhibitory control interprets the modulatory term as a locally decoded error current. Both reject the view that biologically plausible plasticity must be only correlation-based.

5. Realizations in deep networks, spiking systems, and unsupervised representation learning

Error-modulated Hebbian updates are now used in supervised, reinforcement, and unsupervised settings, with different granularities of the third factor: scalar reward, neuron-wise vector, synapse-wise sign, or block-local structural error.

Approach	Setting	Reported result
FFA / Hebbian FFA	MNIST, single-layer, symmetric probability	Analog FFA 95.10%; Hebbian/spiking FFA 92.72%; online Hebbian $\eta>0$7 94.36%
GHL	ImageNet, ResNet-50	Top-1 73.14; Top-5 91.04
Online Hebbian BCI SNN	Zenodo Indy, MC Maze	Pearson $\eta>0$8 and $\eta>0$9; 28–35% memory reduction
SPHeRe	CIFAR-10 / CIFAR-100 / Tiny-ImageNet	81.11% / 56.79% / 40.33%
MOHQA	Hardest POMDPs and Malmo	At least 33% improvement vs. baselines

FFA supplies one explicit bridge from local contrastive learning to neo-Hebbian updates. With squared Euclidean goodness, analog and spiking realizations produced similar accuracy and latent distributions on MNIST, with sparse latent activity vectors having Hoyer Index $y_j$0 and T-SNE separability index $y_j$1. The same paper argues that the forward-only, local nature of the rule makes analog FFA directly relevant for neuromorphic deployment (Terres-Escudero et al., 2024).

In large-scale supervised vision, Global-guided Hebbian Learning (GHL) separates magnitude and direction. The local magnitude is the competitive Oja term

$y_j$2

while direction is provided by the per-synapse sign of the backpropagated gradient,

$y_j$3

The paper reports competitive results on CIFAR-10/100 and ImageNet, including ResNet-50 Top-1 $y_j$4 and Top-5 $y_j$5, and notes that the method maintained performance even for ResNet-1202 on CIFAR-10 (Hua et al., 29 Jan 2026).

Deep networks with asymmetric feedback provide another realization. There the feedforward and feedback weights, $y_j$6 and $y_j$7, are separate, and both are updated locally: $y_j$8 If $y_j$9 is fixed random feedback, performance degrades quickly with depth; if $g_j$0 is learned with the same local rule, performance remains close to ordinary backpropagation, and in the linear case the paper proves that updating the feedback weights accelerates convergence of the error to zero (Amit, 2018).

A complementary engineering route uses surrogate losses whose gradients are exactly Hebbian. With

$g_j$1

appropriate definitions of $g_j$2 recover plain Hebbian, Grossberg’s instar, or Oja updates under autograd. A detached modulatory signal $g_j$3 can be inserted as

$g_j$4

which yields a three-factor update $g_j$5. This makes error-modulated Hebbian learning compatible with modern convolutional frameworks (Miconi, 2021).

Unsupervised representation learning has also adopted local error modulation. SPHeRe defines

$g_j$6

with the auxiliary structural-projection path producing a local modulatory error

$g_j$7

The paper presents the resulting weight change as an error-modulated Hebbian rule driven by presynaptic activity, postsynaptic activity, and a block-local structural mismatch, and reports state-of-the-art results among unsupervised synaptic-plasticity approaches on CIFAR-10, CIFAR-100, and Tiny-ImageNet (Deng et al., 16 Oct 2025).

6. Limitations, misconceptions, and open problems

A recurring limitation is that each derivation is tied to a specific objective and signal geometry. The FFA equivalence to a neo-Hebbian rule depends on the squared Euclidean goodness $g_j$8; other goodness functions may not admit the same clean factorization. HSIC-based Information Bottleneck learning is sensitive to effective batch size $g_j$9 and kernel bandwidth $\Delta w_{ij}=\eta\,m\,y_j x_i$00; small $\Delta w_{ij}=\eta\,m\,y_j x_i$01 severely degrades performance because the kernel estimates become poor. Quantum probability-flow rules assume a single-flip approximation and require calibration of $\Delta w_{ij}=\eta\,m\,y_j x_i$02, $\Delta w_{ij}=\eta\,m\,y_j x_i$03, and $\Delta w_{ij}=\eta\,m\,y_j x_i$04. Policy-gradient and noise-based rules inherit high-variance estimators, sensitivity to stochasticity, and dependence on eligibility timescales. SPHeRe replaces backpropagated task errors with local Gram-matrix mismatch, but this introduces $\Delta w_{ij}=\eta\,m\,y_j x_i$05 batch coupling and reported diminishing gains with depth (Terres-Escudero et al., 2024, Daruwalla et al., 2021, Ohzeki, 1 Jun 2026, Bartlett et al., 2019, Fernández et al., 31 Mar 2025, Deng et al., 16 Oct 2025).

A common misconception is that every error-modulated Hebbian method is fully backpropagation-free. GHL explicitly obtains $\Delta w_{ij}=\eta\,m\,y_j x_i$06 through standard backpropagation; asymmetric-feedback learning still requires backward error activities, even though they travel through separate learned feedback weights; and autograd-based Hebbian CNNs rely on conventional differentiation to realize local rules. Conversely, some biologically motivated methods remain highly structured: proofreading requires synapse-specific validation to avoid catastrophic crosstalk, and dis-inhibitory control assumes a controller and feedback matrices aligned with the network Jacobian (Hua et al., 29 Jan 2026, Amit, 2018, Miconi, 2021, Cox et al., 2010, Rossbroich et al., 2023).

The present literature therefore supports a narrower and more technical reading of the term. Error-modulated Hebbian updates are not a single algorithm but a family of local plasticity decompositions in which pre-post correlation is gated by a third factor. Depending on the model, that factor may be a scalar reward, a vector of layer-local teaching signals, a per-synapse sign, a synapse-specific validation gate, a softmax over instability channels, or a structural mismatch computed in an auxiliary projection. This suggests that the main unifying principle is not any particular biological story or optimization method, but the algebraic separation between local correlation and modulatory control.