# Somato-Dendritic Innovation Learning (SDIL) SDIL is a non-backpropagation learning project motivated by the neuron-specific somato-dendritic residuals reported by Harnett and colleagues. Its central algorithmic hypothesis is narrow: a mixed apical compartment should teach from the component that is unexpected given the same neuron's ordinary somatic state, rather than from raw apical activity. External A6000 collaborators running the complete 81-cell matched crossover should begin with [`COLLABORATOR_ONBOARDING.md`](COLLABORATOR_ONBOARDING.md). It contains the method boundary, current positive and negative results, environment/data setup, frozen matrix, restart policy, and the single entry-point command. For hidden population `l`, the implemented rule is ```text a_l = A_l c + ordinary apical traffic a_hat_l = P_l(h_l), fitted during neutral periods r_l = a_l - a_hat_l Delta W_l = eta (r_l * local postsynaptic gain) h_(l-1)^T. ``` The standard-ResNet stability branch additionally uses a fast paired neutral observation to project any *remaining* affine neutral residual off current soma before plasticity. This local microphase is explicitly counted; it is not described as single-phase or free. `A_l` is calibrated from antithetic node perturbations without reverse-mode differentiation or weight transport. That learned-feedback mechanism is inherited from Lansdell, Prakash, and Kording (2020); it is not claimed as novel. The candidate contribution is the neutral-period, per-cell innovation operation under mixed apical traffic, together with its theory, causal audit, and scaling behavior. See `NOVELTY.md` for the exact prior-art boundary. ## Current audited evidence - On flattened CIFAR-10, SDIL loses only `0.214 +/- 0.349` accuracy points from hidden depth 5 to 60, while DFA's early-layer teaching alignment falls from `0.514` to `0.047`. This is depth preservation on a depth-flat task, not yet evidence that added depth is useful. - With strong soma-predictable apical traffic, raw and norm-matched raw learning fall to `10.38%` and `10.31%`, while innovation learning retains `97.35%`. The paired innovation gain over norm-matched raw is `87.038 +/- 0.607` points across five seeds. - Direct, unamortized node perturbation passes the controlled useful-depth diagnosis, but costs `68.4x` ordinary forward-equivalent work. The learned apical vectorizer, rather than the local eligibility rule, is the current useful-depth bottleneck. - A low-query K1/every-4 calibration protocol retains `112.9%` of the K16 gain over DFA with 16x fewer logical queries, 11x less calibration work, and 5.3x less total forward-equivalent work in the frozen MLP protocol. - The frozen standard-ResNet funnel passed its BP reference and short screen: BP reached `91.62%`, and short-run channel-gated SDIL reached `41.98%` versus tuned DFA's `37.16%`. The full A3 run then failed decisively: SDIL became nonfinite at epoch 89 and ended at chance, while DFA remained finite at `33.06%`. A4 was therefore not opened and no confirmation test seed was touched. - The post-failure representable-subspace estimator passed its exact mechanics tests and improved frozen-forward early/all-layer alignment from `0.0011/0.0117` to `0.0072/0.0527` at matched query count. It nevertheless missed the frozen `0.01` early-layer and improvement gates, so no v2 full ResNet run was launched. - Direct vectorizer-space perturbation then passed its exact mechanics and synthetic variance audit, but V3-1 again failed: early alignment was `0.00714` versus matched V2's `0.00721`, although all-layer alignment rose to `0.06258`. V3 full training therefore remained closed. - Residual response mirroring passes its short gate but fails its sole full run at 10% with NaN validation loss. Its endpoint Q/W cosine nevertheless reaches 0.999998, exposing final alignment as an inadequate certificate for an intermittent tracker. The separate reciprocal KP short gate reaches 82.66%, and its frozen full gate reaches 91.26% versus matched BP's 91.62% with 0.9997 late feedback cosine. The subsequent frozen mixed-traffic screen nevertheless makes raw, norm-matched raw, and innovation all nonfinite in epoch 1. A separately frozen operator-stability branch then rules out fixed predictor margins and passes a dynamic paired-neutral controller: its D1 352-step trajectory is finite, and D2 reaches `83.58%` after 20 epochs versus clean KP's `82.66%` and the failed mixed-traffic controls' `10%`, at `1.3261x` BP MACs and zero task-loss queries. This is still one short validation seed; the predeclared 200-epoch D3 run subsequently passes all 19 gates at `91.18%` versus BP's `91.62%` and clean KP's `91.26%`, with `0.9994` early alignment, zero task-loss queries, and `1.3261x` BP MACs. This raises the strict score to 6/10. The independently frozen paired five-seed D4 test panel then passes every gate: dynamic innovation reaches `91.584%` mean test accuracy versus clean KP's `91.388%`, wins the mean pairing by `0.196` points, and has a clean-minus-dynamic one-sided 95% upper bound of only `0.131` points. Every dynamic seed reaches at least `91.51%`, mean early alignment is `0.999687`, and all trajectory, projection, leakage, query, hardware, MAC, memory, split, and test-isolation checks pass. This untouched confirmation establishes the strict 7/10 accept bar; D4 alone does not establish that added standard-network depth is useful. - Native author-code fidelity is complete. BurstCCN reaches `80.10%` at its validation-selected epoch versus published `82.97 +/- 0.21%`; Dual Prop reaches `92.46%` versus published `92.41 +/- 0.07%`. Their audited walls are `15451.3 s` and `23119.8 s`, respectively, and they are not presented as matched-compute points. The broad endogenous-traffic gate and the Harnett desired-velocity/online-control screen failed. Those results are retained and explicitly constrain the paper: SDIL does not currently explain the complete reported dendritic population signature or online control, and the residual mechanism is supported for traffic predictable from the chosen somatic statistic rather than arbitrary top-down context. `ROADMAP.md` and `ORAL_B.md` contain the frozen gates and complete negative branches. A temporal-difference recovery separates perturbation-learned causal role from within-episode performance velocity. Its development gate selects `eta=0.1` with 98.05% worst-task success, and its untouched 30-record confirmation retains 99.53% mean success, positive sign inversion in all 30 records, and a 0.639 velocity-over-error correlation advantage. The complete R2 gate nevertheless fails: residual outcome decoding is only 47.33%, residuals trail soma by 4.32 points, and the longitudinal correlation is `-0.013`. Thus the recovery sharpens the boundary--causal-role temporal-difference plasticity works in this synthetic task, but the broader Harnett-like population vectorization signature is not established. At that stage the strict score remained 7/10 and the old oral-A depth panel stayed sealed. An independent oral-B-v2 then adds an explicit terminal reward/timeout phase, a local linear TD critic, and temporal eligibility traces. Its first frozen grid is retained as a cold-start failure, and a fixed-target recovery is retained after one new seed remains at ceiling. The final algorithm is frozen before a label-free psychometric calibration: cursor-max quantiles from 512 outcome-free trials define targets for independent rewarded/timeout trials. The complete untouched 6-task by 5-model confirmation passes every clustered gate: 100% task success, 0.976 mean learned-role cosine, 0.063 residual--soma correlation, 54.37% surrounding-state decoding, 50.09% rewarded trials, 99.83% terminal outcome decoding, a 0.400 acute outcome-lesion drop in role separation, and 30/30 positive causal signs. This raises the formal milestone to 8/10 and permits only a separately frozen oral-A-v2 protocol; the old depth panel remains closed. That separately frozen oral-A-v2 protocol is now complete. Its audited 60 records cross BP, tuned DFA, clean reciprocal KP, and dynamic SDIL over ResNet-20/32/56 with five paired seeds. Dynamic SDIL rises `91.584% -> 92.254% -> 92.760%`; all five depth-20-to-56 pairs improve, with a mean gain of `1.176` points. ResNet-56 SDIL reaches `92.760%` versus `92.632%` BP, `92.670%` clean KP, and `30.850%` DFA while retaining `0.999423` early-third alignment at `1.331x` the matched BP MAC estimate. This establishes the internal 9/10 standard-depth milestone. It does not erase the failed old branch or show that residualization, rather than the inherited reciprocal backbone, causes clean-task scaling. The next matched crossover is mechanically registered as nine distinct architecture/size points times nine methods: miniCNN/VGGlike/VGG16, ResNet-20/32/56, and decoder-Transformer-4/8/12, each with BP, ordinary FA, DFA, PEPITA, Forward--Forward, EP, Dual Propagation, clean KP, and SDIL. All 81 cells are mandatory; adding a width, context, or depth point adds all nine methods rather than an SDIL-only extension. The complete Plain-CNN P2 panel now passes its 27/27 audit. SDIL scales from `82.90%` to `89.58%` to `90.70%` validation accuracy from miniCNN through VGG16, while ordinary FA falls from `68.14%` to `33.80%` and DFA becomes nonfinite at VGG16. Dual Propagation reaches `92.38%` on VGG16 but takes `5.97` hours versus SDIL's `1.24`; clean KP slightly dominates SDIL there at `90.86%` in `1.08` hours. Thus the first family supports scaling and a cost advantage over iterative strong baselines, but not global Pareto dominance. ResNet and Transformer crossover cells remain gated on their complete validation-only selectors. ## Publication-facing artifacts - `RESULTS.md`: audited positive and negative results; - `COLLABORATOR_ONBOARDING.md`: complete A6000 81-cell replication packet; - `THEORY.md`: estimator variance, descent conditions, conditional innovation, timescales, and hardware-independent cost; - `BASELINES.md`: matched and native-author baseline ledger; - `ROADMAP.md`: accept/oral evidence gates and their state; - `MIXED_TRAFFIC.md`: frozen standard-ResNet raw/matched/innovation accept gate; - `DYNAMIC_INNOVATION.md`: post-failure two-timescale stability theory and frozen D1--D4 gates; - `ORAL_B_RECOVERY.md`: structural diagnosis and mechanics-only temporal- difference recovery boundary; - `ORAL_B_V2.md`, `ORAL_B_V2_RECOVERY.md`, and `ORAL_B_V2_CALIBRATED_RECOVERY.md`: retained v2 failures and the passed label-free calibrated outcome-surprise protocol; - `ORAL_A.md`: frozen standard CIFAR ResNet funnel; - `ORAL_A_V2.md`: frozen post-failure representable-subspace funnel; - `ORAL_A_V3.md`: frozen vectorizer-space causal-calibration funnel; - `ORAL_A_RECOVERY_V2.md`: passed calibrated-BCI-gated ResNet-20/32/56 scaling protocol; - `CROSS_ARCHITECTURE_CROSSOVER.md`, `RESNET_CROSSOVER.md`, and `TRANSFORMER_CROSSOVER.md`: frozen 81-cell matched-crossover contracts; - `REVIEW_SCORECARD.md`: adversarial ICLR-style score trajectory; - `paper/MANUSCRIPT.md`: evidence-bound ICLR working draft; - `paper/CLAIM_LEDGER.json` and `paper/manuscript_audit.json`: direct bindings from manuscript numbers, gate statuses, figures, and claim boundaries to their audited source files; - `results/figs/`: deterministic PDF/PNG main figures, captions, and a source hash manifest, including the untouched D4 ResNet-20, oral-B-v2, and 60-record standard-depth confirmations, plus audited RRM failure and dynamic-stability supplements. The current main figures show the local-method Pareto frontier, credit assignment versus depth, the load-bearing innovation ablation, and the untouched standard-ResNet confirmations, including positive ResNet-20-to-56 scaling. The audited complete Plain-CNN crossover is rendered separately in `results/figs/figure7_plain_cnn_pareto.{pdf,png}`, with local-only and BP-inclusive frontiers kept distinct. The first three main figures require exactly seeds 0--4; the ResNet panels require paired test seeds 10--14. Strict renderers refuse missing, dirty, protocol-mixed, source-drifted, or gate-inconsistent cells. ## Verification The lab environment currently used for audited CPU checks is `/home/yurenh2/miniconda3/envs/ep_pascal`. Run: ```bash experiments/finalize_accept.sh ``` This regenerates all publication-facing figures and manifests, enforces the frozen Pareto/scaling, D4, oral-B-v2, and 60-record standard-depth gates, verifies baseline and protocol mechanics, checks the local-rule smoke tests, evaluates the theory identities, and rebuilds the audited tables. The convolutional infrastructure can be checked independently: ```bash /home/yurenh2/miniconda3/envs/ep_pascal/bin/python3 experiments/cifar_image_smoke.py /home/yurenh2/miniconda3/envs/ep_pascal/bin/python3 experiments/conv_local_smoke.py ``` These checks cover deterministic CIFAR splitting/augmentation, the standard CIFAR `6n+2` ResNet topology, exact local convolution and BatchNorm eligibilities, the BatchNorm-coupled perturbation objective, predictor fitting, translation-shared feedback, representable-subspace causal calibration, and parameter/cost accounting. The latter is a transparent post-A3 development path: it perturbs the channel-gated basis directly and reproduces the expected full-field delta rule with far less irrelevant spatial variance. It does not retroactively reopen or relabel the failed frozen Oral-A protocol. The same smoke suite also audits a convolutional hierarchical-FA baseline. Its random feedback tensors follow the real residual DAG and use local activation Jacobians, but remain independent of forward weights. An audit-only symmetric copy reproduces exact hidden gradients to `5.3e-16` relative error and the BP parameter update to `2.98e-8`; actual HFA never performs that copy. `HFA_BASELINE.md` freezes its matched ResNet-20 validation screen before any HFA accuracy endpoint is observed. The learned-hierarchy development path perturbs the feedback convolution parameter spaces themselves. One antithetic pair estimates every edge's causal target moment simultaneously, while the current prediction moment is a local child/parent correlation. Its BatchNorm-coupled JVP and symmetric-limit delta rule are executable smoke tests. `ORAL_A_V4.md` freezes the causal-capture and conditional accuracy gates; no accuracy claim is attached until they pass. Because the global-scalar V4 gate failed, the next strong inherited comparator uses normalized local response mirroring. Its observation phase sends Gaussian parent activity through ordinary local forward synapses; a separate feedback update sees only the probe/child-response pairs. The smoke suite verifies both high finite-sample kernel recovery and that changing forward parameters after observation cannot affect the update. Any scale obtained this way is credited to the weight-mirror baseline until innovation residualization is shown to be load-bearing on top of it. `MIRROR_BASELINE.md` freezes that comparator's CIFAR-prefix capture, short accuracy, and conditional full-validation gates. After estimate-then-average WM narrowly failed its short accuracy gate, the next substantive inherited control uses residual response mirroring. Q predicts the observed child response and learns only from the local prediction error; unlike WM, its update is zero for every probe at exact symmetry. The extra feedback prediction convolution is charged explicitly, and this rule receives no SDIL novelty credit. `RESIDUAL_MIRROR.md` freezes its capture and conditional task gates. The stronger reciprocal Kolen--Pollack substrate is audited separately in `KP_BASELINE.md`. Its short and full frozen gates pass; the full record reaches 91.26% validation accuracy, 0.9994 early alignment, and 1.326x BP MACs without feedback loss queries. Before that full endpoint, `MIXED_TRAFFIC.md` froze the actual Harnett-specific test: raw, norm-matched raw, and innovation under identical four-times-RMS soma-predictable apical traffic and predictor cost. Mechanics pass, but the complete MT-1 panel fails: all three signals become nonfinite in epoch 1 and end at 10.00% validation accuracy. MT-2 and MT-3 remain untouched, closing this standard-ResNet recovery without a weaker traffic intervention. The subsequent V3 mechanism estimates the required A/G matrix statistics directly by perturbing the vectorizer parameter subspace. It remains forward-only and uses two causal loss queries per direction. Its exact and BatchNorm JVP identities, delta rule, and matched synthetic variance comparison are covered by `experiments/conv_local_smoke.py`; no task-level claim is made until a new frozen gate is completed. `experiments/finalize_accept.sh` additionally requires strict imports of the frozen BurstCCN and Dual Propagation author-code runs. Both records now pass; the accept-bar mechanical audit is green. The original standard-ResNet branch stopped at its failed A3 validation gate and its A4 panel remains sealed; the independently authorized v2 recovery is the separate passed 60-record panel. ## Result discipline Every publication-facing run records the source revision, tracked-tree state, full arguments, split identity, seed, evaluation count, wall time, query/MAC accounting, and final finiteness. Development, validation, and untouched confirmation results are never pooled. Failed gates close their branch instead of triggering seed deletion or post-hoc threshold changes. The current formal milestone is `9/10` after the untouched D4, oral-B-v2, and standard-depth confirmations. A conservative external-review forecast is `8/10`: the full matched nine-method crossover, original-data biological validation, and a credit pathway novel beyond inherited perturbation/KP mechanisms remain open. Scores change only after an audited frozen stage, not after a pilot or presentation improvement.