{"id":"9f12be1a-fe41-4271-b2be-f1c5c46b59c2","arxiv_id":"2607.10302","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Direct bidiagonal-based methods (CK deflation, dual normals, inplace, Lanczos reinterpretation) compute Moore-Penrose and group pseudoinverses while mitigating conditioning, rank, and sparsity issues for large-scale regression.","lead":"Direct methods can compute Moore-Penrose pseudoinverses of matrices without full SVD or eigenvalue work, by reducing to bidiagonal forms and handling sparse cases via block tearing or Lanczos reinterpretation. This matters for large-scale sparse multilinear regression in data science and ML where dense pseudoinverses or conditioning blowups are bottlenecks.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified beyond the reader's already-flagged sparse-validation gap.","rationale":"The reader's weakest_assumption correctly identifies the load-bearing practical premise (bidiagonalization remains usable and sparsity-preserving). The mathematical development (Theorems 1–3, Propositions, Woodbury reduction) holds under that premise, code is supplied, and no stronger internal flaw is present. Therefore the CONDITIONAL verdict already reflects the right residual risk; no adjustment is required.","tokens_in":17850,"tokens_out":398,"duration_ms":9678,"concrete_test":"On a single large sparse test matrix (e.g., 10^5–10^6 nonzeros from a multilinear-regression design matrix), time and memory-profile Method 3 (inplace + Cython gebrd) against scipy.linalg.pinv and LSQR for a batch of right-hand sides; if peak memory exceeds O(nnz) by more than a small constant or wall time is not competitive, the practicality claim for sparse regimes weakens.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Abstract, §1, Theorems 1–3, §5) is that the three bidiagonal routes (CK deflation, dual-normal Thomas, inplace blocks) plus the Lanczos rank-one reinterpretation overcome conditioning, rank-revealing, and densification limitations for large/sparse multilinear regression. The theorems are correctly proved under the stated assumptions, the Woodbury simplification (Eq. 11) is algebraically valid for the tear geometry, and the GitHub implementations exist. The only soft spot is the same one the reader already isolates: that full/partial bidiagonalization plus O(n) bidiagonal work remains practical and sparsity-preserving at the scales claimed. No deeper internal inconsistency, hidden rank assumption, or conditioning failure appears in the argument itself.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The paper argues that direct (non-SVD) methods for the Moore-Penrose and group pseudoinverses can overcome the classic practical limitations of conditioning (squared condition numbers of AᵀA/AAT), rank-revealing requirements, and densification of sparse matrices. It reduces the problem to bidiagonal matrices via Householder or Lanczos bidiagonalization, then supplies three constructive routes for the bidiagonal pseudoinverse (CK deflation by Givens chasing into invertible C and superdiagonal K, dual-normal Thomas solves on the block-tridiagonal BBᵀ, and explicit in-place formulas for the three possible square/rectangular blocks after splitting on zero etaⱼ and etaⱼ=eta_{j}=0). A Lanczos reinterpretation as successive rank-one updates that produce a weak (1,2,3) pseudoinverse is also given, together with a complex-regularization limit that recovers the group inverse of the Jordan-Wielandt augmentation. Theorems 1–3, the simplified Woodbury tear (11), and competitive timings versus scipy.linalg.pinv (Table 2) are offered as evidence that the methods are suitable for large-scale or sparse multilinear regression.","tokens_in":18014,"tokens_out":1062,"duration_ms":11286,"significance":"If the practical claims hold, the work supplies a useful algorithmic toolkit for applications that repeatedly apply a pseudoinverse (or a weak pseudoinverse) to many right-hand sides and that cannot afford a full SVD or a dense matrix. The explicit block formulas, the matrix-free back-substitution realization of C^{-1}, the Woodbury reduction of storage to O(r), and the publicly available Cython/LAPACK implementations are concrete engineering contributions. The reduction of the Moore-Penrose case to a group inverse of a sparse symmetric augmentation is also a clean conceptual observation that may seed further regularization algorithms. The paper therefore has clear value for numerical linear algebra and sparse least-squares practice even if the large-scale sparse validation remains incomplete.","major_comments":[{"comment":"§5 and the Abstract claim that the three bidiagonal routes (plus Woodbury tearing (11)) are appropriate for large-scale or sparse multilinear regression and that they avoid densification. The only numerical evidence is Table 2 (dense m,n ≤ 1000 versus scipy.linalg.pinv). No sparse matrix suite, fill-in statistics for gebrd/orgbr or Lanczos, memory profiles, or comparison against LSQR/sparse direct solvers is supplied. Without such evidence the central practical claim remains unsubstantiated; at minimum a sparse benchmark section (or a clear restriction of the claim to the dense/bidiagonal regime) is required.","section":null},{"comment":"§4, Theorem 3 constructs a weak (1,2,3) pseudoinverse by successive rank-one updates, then notes that condition 4 can be restored by a post-processing formula involving an arbitrary Z. For the least-squares applications emphasized in §1–2 (Eqs. (1)–(2)), the weak inverse already suffices, but the manuscript never states this limitation cleanly nor quantifies the extra cost or numerical effect of restoring the fourth Penrose condition when it is needed. A short clarifying paragraph and, if possible, a numerical illustration would remove the ambiguity.","section":null}],"minor_comments":[{"comment":"Figure 1 is typeset with many unreadable control characters and overlapping fractions; it should be regenerated as a clean LaTeX array or TikZ diagram.","section":null},{"comment":"§3, after Theorem 1: the claim that “the study of pseudoinverses can be reduced to the study of pseudoinverses of bidiagonal matrices” is slightly overstated for the Drazin case (index >1); a one-sentence qualification would be accurate.","section":null},{"comment":"Table 2 caption and surrounding text: report absolute wall-clock times (or FLOPs) in addition to the relative factors so that readers can judge absolute performance.","section":null},{"comment":"Several bibliographic entries (e.g., the ICML 2025 Schwarz-Schur paper) appear to be future or arXiv-only; verify publication status and supply stable identifiers.","section":null},{"comment":"Notation: both A^i and A_i are used for pseudoinverses of different types; a short notational table would help.","section":null}],"recommendation":"major_revision","confidential_remarks":"The mathematical core (Theorems 1–3, Woodbury simplification, group-inverse reduction) is sound and the GitHub code is a genuine plus. The manuscript’s main weakness is the gap between the sparse/large-scale rhetoric and the dense-only experiments; once that is closed or the claims are narrowed, the paper becomes a solid methods contribution suitable for a numerical-linear-algebra venue. No integrity or novelty-disclosure concerns."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful takeaway is that Knisley reinterprets Golub-Kahan/Lanczos bidiagonalization, LSQR-style rank-one updates, MacDuffee, and Woodbury tearing into three concrete direct routes (CK deflation, dual-normal Thomas, inplace double-split blocks) plus a complex-regularization view of the group inverse of the Jordan-Wielandt matrix. These are aimed squarely at sparse multilinear regression where you want to avoid squared condition numbers, exact rank revelation, and dense A† storage.\n\nWhat is new is the packaging and the explicit formulas. Theorem 1 (Givens chasing to C ⊕ K) is standard but cleanly stated; Proposition 1 and the three block types give closed-form inplace inverses; Theorem 2–3 show that the Lanczos residual R_k \to 0 implies the Penrose residual vanishes and that successive rank-one updates produce a weak (1,2,3) pseudoinverse. The Woodbury simplification for bidiagonal tears (Vᵀ B̂^{-1} U = 0) is algebraically correct and keeps memory O(r). The complex-regularization identity recovering A# from Re((A + iεI)^{-1}) is a nice observation. GitHub code and Table 2 timings (Method 3 competitive with scipy.linalg.pinv) are real assets.\n\nSoft spots are modest and already flagged by the reader. Full or partial bidiagonalization is still the bottleneck; the paper asserts that the subsequent O(n) or O(n^{2}) work plus matrix-free back-substitution preserves sparsity for “large” problems, but the numerical evidence is small-scale and dense-matrix oriented. Conditioning of the dual-normal route is acknowledged but not stress-tested on ill-conditioned sparse examples. Prose is dense and the organization jumps between MP, Drazin, and group inverses. None of these break the math.\n\nThis is for numerical linear algebraists and practitioners who already use LSQR or sparse least-squares and want a direct, reusable pseudoinverse operator. The citation pattern is appropriate (Golub–Kahan, Paige–Saunders, Ben-Israel–Greville, Higham). I would send it to peer review; a numerical journal referee can demand larger sparse benchmarks and tighter complexity statements, but the core is sound enough to deserve that time. Engage if you care about sparse regression kernels; otherwise skim the theorems and the GitHub.","headline":"Solid packaging of bidiagonal direct routes for the MP (and group) pseudoinverse; theorems hold, code exists, sparse-scale claims still under-demonstrated.","tokens_in":18625,"tokens_out":608,"would_cite":false,"duration_ms":6858,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["15A09","65F20","65F50"],"pacs":[],"model":"grok-4.5","headline":"Direct methods for Moore-Penrose pseudoinverses overcome conditioning, rank, and densification barriers and work for large sparse multilinear regression.","keywords":["pseudoinverses","regression","sparsity","bidiagonalization","Moore-Penrose","Lanczos","group inverse","least squares"],"falsifier":"Time and residual-accuracy comparisons of the three bidiagonal methods against SVD-based pinv on a suite of sparse regression matrices of growing size and density; if the direct methods lose both speed and accuracy once the matrices become realistically sparse, the practicality claim fails.","tokens_in":18733,"feed_emoji":"🔢","tokens_out":859,"duration_ms":18929,"temperature":0.7,"pith_summary":"The paper shows that direct methods for the Moore-Penrose pseudoinverse, long dismissed as rank-revealing, ill-conditioned, or density-creating, can be made practical for large-scale and sparse least-squares problems. The strategy reduces any matrix to a bidiagonal factor via Householder or Lanczos processes, then inverts that bidiagonal by one of three explicit techniques: CK deflation into invertible and nilpotent blocks, dual-normal tridiagonal solves, or in-place formulas on the natural zero-split blocks. A fourth construction reinterprets Lanczos itself as successive rank-one updates that converge to a weak pseudoinverse. Because the expensive work stays on the sparse bidiagonal and Woodbury tearing keeps storage linear, the resulting pseudoinverse can be applied repeatedly without ever forming a dense matrix. Readers who care about multilinear regression on sparse data therefore gain a family of direct solvers that sit between full SVD and pure iterative methods.","feed_headline":"Direct pseudoinverses work for large sparse regression","feed_subtitle":"They dodge squared condition numbers, exact rank, and dense fill-in by operating on bidiagonals","key_machinery":"Bidiagonal deflation (Theorem 1) that splits an m\times(m+1) bidiagonal into an invertible bidiagonal C and a superdiagonal K, together with the three concrete solvers (CK, dual-normal Thomas, inplace blocks) and the Lanczos rank-one update construction of Theorem 3.","core_discovery":"Direct methods applied after bidiagonalization overcome the three classic limitations of conditioning (squared condition numbers of AᵀA), exact rank revelation, and densification of sparse matrices, and are therefore appropriate for large-scale or sparse multilinear regression.","pith_inferences":["The block formulas and Woodbury tearing are natural candidates for hardware-level sparse tensor cores already used in large language models.","The same CK and inplace machinery can be specialized to graph-Laplacian or attention-mask sparsity patterns that appear in modern transformers.","Because the methods only require existing LAPACK bidiagonalization kernels, they can be dropped into production libraries with modest engineering cost."],"forward_implications":["A sparse matrix can keep a matrix-free or low-memory pseudoinverse that never densifies.","Multilinear problems with many right-hand sides can form the pseudoinverse once and reuse it.","Woodbury tearing of the bidiagonal reduces inverse storage from quadratic to linear.","The same bidiagonal toolkit yields both Moore-Penrose and group inverses via the Jordan-Wielandt embedding.","Complex regularization Re((A+iεI)^{-1}) supplies a rapidly converging family of approximations without squared condition numbers."],"fun_headline_variants":["Direct methods after bidiagonalization beat pseudoinverse limits","Bidiagonalization frees direct pseudoinverses for sparse regression","Direct pseudoinverses sidestep rank conditioning and fill-in","Bidiagonal direct methods suit large sparse multilinear regression","Reinterpreted direct methods overcome classic pseudoinverse barriers"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The methods assume that bidiagonalizing the original large sparse matrix remains practical and does not destroy the sparsity that makes the later O(n) or O(n^{2}) work advantageous.","fun_headline_variants_meta":{"raw":{"variants":["Direct methods after bidiagonalization beat pseudoinverse limits","Bidiagonalization frees direct pseudoinverses for sparse regression","Direct pseudoinverses sidestep rank conditioning and fill-in","Bidiagonal direct methods suit large sparse multilinear regression","Reinterpreted direct methods overcome classic pseudoinverse barriers"]},"model":"grok-4.5","effort":"low","cost_usd":0.003884,"raw_usage":{"total_tokens":1072,"prompt_tokens":589,"num_sources_used":0,"completion_tokens":83,"cost_in_usd_ticks":38840000,"prompt_tokens_details":{"text_tokens":589,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":400,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":589,"tokens_out":83,"duration_ms":4082,"temperature":1.0,"reasoning_tokens":400,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-14T12:46:36.291289+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Time and residual-accuracy comparisons of the three bidiagonal methods against SVD-based pinv on a suite of sparse regression matrices of growing size and density; if the direct methods lose both speed and accuracy once the matrices become realistically sparse, the practicality claim fails.","supporting_citations":[],"review_version":1}