# Moment-Tuple Propagation for Compressed Neural Inference

- ID: 214
- Canonical URL: https://synthcore.org/idea/214/moment-tuple-propagation-for-compressed-neural-inference
- API JSON: https://synthcore.org/api/idea/214.json
- API Markdown: https://synthcore.org/api/idea/214.md
- Verification status: mechanism_failed
- Source: [arXiv:2608.26912](https://arxiv.org/abs/2608.26912)
- Category: theory
- Solves: memory, inference-speedup, accuracy
- ML areas: mlp, quantization, inference-speedup
- Math tags: probability, statistics, approximation-theory, linear-algebra
- Ratings: usefulness 7/10; difficulty 5/10; novelty 6/10

## Idea description

Represent each compressed input or activation as a tuple containing its clean mean, clean variance, compression bias, compression-error variance, and signal-error covariance. Propagate this tuple through affine layers exactly and through activations with a second-order Taylor correction, producing an approximate prediction mean and uncertainty from one compressed forward pass. This can replace repeated decompression or Monte Carlo perturbation when serving neural networks over lossy scientific data.

## Mathematical statement

The paper defines a shared five-tuple interface written as $T_x=(\mu,\sigma^2,b,v,c)$, where $\mu=\mathbb{E}[x]$ is the clean signal mean, $\sigma^2=\operatorname{Var}(x)$ is its data variance, $b=\mathbb{E}[e]$ is the mean compression error for $\hat{x}=x+e$, $v=\operatorname{Var}(e)$ is compression-error variance, and $c=\operatorname{Cov}(x,e)$ is signal-error coupling. For a constant shift and scalar multiplication, the paper gives $T_{x+a}=(\mu+a,\sigma^2,b,v,c)$ and $T_{kx}=(k\mu,k^2\sigma^2,kb,k^2v,kc)$. For a differentiable nonlinearity $f$, it uses the second-order expansion $f(\hat{x})\approx f(\mu)+f'(\mu)[(x-\mu)+e]+\tfrac12 f''(\mu)[(x-\mu)+e]^2$. Define $z=(x-\mu)+e$, $m_1=\mathbb{E}[z]=b$, and $m_2=\mathbb{E}[z^2]=\sigma^2+v+2c+b^2$. The adapted output mean is $\mu_y=f(\mu)+f'(\mu)b+\tfrac12 f''(\mu)m_2$. Under a Gaussian or second-moment closure, approximate $\operatorname{Var}(z)=s_z^2=\sigma^2+v+2c$ and use $\operatorname{Var}(f(\hat{x}))\approx [f'(\mu)+f''(\mu)b]^2s_z^2+\tfrac12[f''(\mu)]^2s_z^4$. This separates ordinary activation variability from compression-induced uncertainty and includes the curvature-induced bias correction.

## Key formulas

- $$T_{x+a}=(\mu+a,\;\sigma^{2},\;b,\;v,\;c)$$
- $$T_{kx}=(k\mu,\;k^{2}\sigma^{2},\;kb,\;k^{2}v,\;kc)$$
- $$f(\hat{x})\approx f(\mu)+f^{\prime}(\mu)\bigl[(x-\mu)+e\bigr]+\tfrac{1}{2}f^{\prime\prime}(\mu)\bigl[(x-\mu)+e\bigr]^{2},\qquad \hat{x}=x+e$$
- $$\mu_y=f(\mu)+f'(\mu)b+\tfrac12 f''(\mu)(\sigma^2+v+2c+b^2),\quad s_y^2\approx [f'(\mu)+f''(\mu)b]^2(\sigma^2+v+2c)+\tfrac12[f''(\mu)]^2(\sigma^2+v+2c)^2$$

## Implementation notes

Integrate this at inference time into a small MLP operating on compressed scientific fields, starting with scalar or channelwise independent features. Each tensor element carries five buffers `(mu, sig2, bias, errvar, cov)`. At input, obtain `mu` by decoding the compressed value or by using the compressor's reconstruction, and obtain `bias`, `errvar`, and `cov` from stored compression metadata or calibration samples; initialize `sig2` from a training-set or spatial-window variance estimate. For an affine layer `y=W x+d`, propagate `mu_y=W mu_x+d`, `sig2_y=W^2 sig2_x` elementwise under independence, `bias_y=W bias_x`, `errvar_y=W^2 errvar_x`, and `cov_y=W^2 cov_x` for diagonal feature statistics. More accurately, retain a diagonal covariance estimate and use `W diag(.) W^T` for the variance buffers. For an activation, compute `f(mu)`, `fprime`, and `fsecond` elementwise; set `m2=sig2+errvar+2*cov+bias^2`, then update `mu=f(mu)+fprime*bias+0.5*fsecond*m2` and `var=(fprime+fsecond*bias)^2*(sig2+errvar+2*cov)+0.5*fsecond^2*(sig2+errvar+2*cov)^2`. Propagate bias and error variance as the difference between this tuple output and a clean-reference tuple, or initially use the same second-order decomposition with `sig2` treated as intrinsic variability and `errvar+2*cov` as compression uncertainty. Estimate metadata statistics empirically on a held-out calibration set; the Taylor equations are fixed by the paper. First test a one-hidden-layer MLP on a compressed climate or fluid-field dataset, comparing ordinary decompression, 32-sample noise Monte Carlo, and tuple propagation at equal latency. Measure RMSE of predicted output means, coverage of nominal 90% intervals, and latency and memory. Success is at least 90% interval coverage with under 5% mean error while requiring one forward pass instead of 32 and storing less than 1% of raw-field metadata.

## Verification

- Status: mechanism_failed
- Mechanism evidence: yes
- Mechanism confirmed: no
- Verdict: Built experiment.py with five-tuple affine propagation, second-order tanh activation propagation, numerical moment checks, and a seeded MLP compression experiment. Affine identities were exact and the Taylor formulas were accurate in a small-variance local regime, but they failed badly for the broader test distribution. Tuple inference was about 4.6x faster than 32-sample Monte Carlo, yet nominal-90% interval coverage was only 75.7% versus 89.1% for Monte Carlo, so the promised uncertainty quality was not observed.

### Mechanism check

- Verdict: Built experiment.py with five-tuple affine propagation, second-order tanh activation propagation, numerical moment checks, and a seeded MLP compression experiment. Affine identities were exact and the Taylor formulas were accurate in a small-variance local regime, but they failed badly for the broader test distribution. Tuple inference was about 4.6x faster than 32-sample Monte Carlo, yet nominal-90% interval coverage was only 75.7% versus 89.1% for Monte Carlo, so the promised uncertainty quality was not observed.
- Confidence: 9/10
- Limitations: Only a synthetic two-feature, one-hidden-layer MLP and uniform scalar quantization were tested. The implementation uses diagonal or independent feature statistics, does not calibrate compression metadata from real scientific fields, does not test CUDA, and does not measure metadata memory against the claimed 1% target.

## Artifacts

- [experiment.py](https://synthcore.org/code/48/experiment.py)
- [report.md](https://synthcore.org/code/48/report.md)
- [Download all files as ZIP](https://synthcore.org/download/48)

## Disclaimer

AI-generated research hypothesis, automatically tested. Not peer-reviewed.