feat: GF-T on-chip binary logistic classifier (proven on AX7203, 100% held-out) - #1821
Merged
Conversation
gft_logistic.t27: on-chip SGD step of a binary classifier -- logit z=w0*x0+w1*x1, p=hard_sigmoid(z)=clamp(0.5+0.25*z,0,1), gradient dL/dz=p-y, update w_j'=w_j-eta*(p-y)*x_j; returns (w0'<<32)|w1'. Uses a division-free hard-sigmoid because a runtime reciprocal maps to a $div/CARRY4 the open P&R flow cannot place; only smul/sadd/compares. Proven on a live AX7203 (uart_logistic.v): streaming labeled 2D points for a hidden boundary (class1 iff x0+x1>0), the board learns the weights on-chip and classifies 8/8 held-out points correctly -- 100% generalization. Classification, not just regression. In-spec tests PASS. docs/NOW.md updated. Refs #1764 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
New spec gft_logistic.t27 — on-chip SGD step of a binary classifier: logit z=w0x0+w1x1, p=hard_sigmoid(z)=clamp(0.5+0.25z,0,1), gradient dL/dz=p-y, update w_j-eta(p-y)*x_j; returns (w0'<<32)|w1'. Division-free hard-sigmoid (a runtime reciprocal maps to a $div/CARRY4 the open P&R flow cannot place). In-spec tests PASS. Proven on a live AX7203 (uart_logistic.v): streaming labeled 2D points for a hidden boundary (class1 iff x0+x1>0), the board learns on-chip and classifies 8/8 held-out correctly (100% generalization) — classification, not just regression. docs/NOW.md updated. Refs #1764