Contraction-Gauge Preconditioning for Quantized Matrix Multiplication

We study low-precision computation of C=AB with both factors quantized. We derive an exact finite-dimensional identity for the expected squared product error under independent, zero-mean entrywise errors with known variance fields; it holds exactly for non-overloading subtractive dither and for independent stochastic rounding, and we empirically assess deterministic round-to-nearest (RTN). Using the product-preserving equivalence AB=(AT)(T^{-1}B), we formulate contraction-gauge preconditioning: jointly choosing a factor representation and its sharing pattern before quantization. Preconditioning can reduce product error but may require extra transformed, quantized copies of the opposite operand: a shared transform needs one copy, a block-specific transform up to one per block. Within the bounded family of positive diagonal gauges (folds), a geometric program computes a globally optimal shared fold and a linear program decides whether the identity fold is already optimal. For other families we derive computable selection statistics -- tail index for scaling, profile spread for partitioning, coherence and weighted-Gram energy for rotations, slice-energy covariance for hierarchy depth -- with upper bounds for ranking heuristic candidates. Across twelve linear products from a trained three-block image classifier, median within-product rank correlations between dither-model predictions and deterministic-RTN errors are 0.937 at 8 bits and 0.918 at 4 bits. The GP fold cuts held-out product error over the identity fold by 18.0% (8-bit) and 20.5% (4-bit) in geometric mean, beats a SmoothQuant-style grid baseline at both precisions and on ten of twelve products, and lowers composed logit MSE by 15.4% and 26.4%. We thus provide exact stochastic product-error accounting, certified selection within the diagonal family, and a common objective for evaluating reusable transform candidates under RTN.

Ghosts of Softmax: Complex Singularities That Limit Safe Step Sizes in Cross-Entropy

Optimization analyses for cross-entropy training rely on local Taylor models of the loss to predict whether a proposed step will decrease the objective. These surrogates are reliable only inside the Taylor convergence radius of the true loss along the update direction. That radius is set not by real-line curvature alone but by the nearest complex singularity. For cross-entropy, the softmax partition function $F=\sum_j \exp(z_j)$ has complex zeros -- ``ghosts of softmax'' -- that induce logarithmic singularities in the loss and cap this radius. To make this geometry usable, we derive closed-form expressions under logit linearization along the proposed update direction. In the binary case, the exact radius is $ρ^*=\sqrt{δ^2+ π^2}/Δ_a$. In the multiclass case, we obtain the lower bound $ρ_a=π/Δ_a$, where $Δ_a=\max_k a_k-\min_k a_k$ is the spread of directional logit derivatives $a_k=\nabla z_k\cdot v$. This bound costs one Jacobian-vector product and reveals what makes a step fragile: samples that are both near a decision flip and highly sensitive to the proposed direction tighten the radius. The normalized step size $r=τ/ρ_a$ separates safe from dangerous updates. Across six tested architectures and multiple step directions, no model fails for $r<1$, yet collapse appears once $r\ge 1$. Temperature scaling confirms the mechanism: normalizing by $ρ_a$ shrinks the onset-threshold spread from standard deviation $0.992$ to $0.164$. A controller that enforces $τ\leρ_a$ survives learning-rate spikes up to $10{,} 000\times$ in our tests, where gradient clipping still collapses. Together, these results identify a geometric constraint on cross-entropy optimization that operates through Taylor convergence rather than Hessian curvature.

What Trace Powers Reveal About Log-Determinants: Closed-Form Estimators, Certificates, and Failure Modes

Computing $\log\det(A)$ for large symmetric positive definite matrices arises in Gaussian process inference and Bayesian model comparison. Standard methods combine matrix-vector products with polynomial approximations. We study a different model: access to trace powers $p_k = \tr(A^k)$, natural when matrix powers are available. Classical moment-based approximations Taylor-expand $\log(λ)$ around the arithmetic mean. This requires $|λ- \AM| < \AM$ and diverges when $κ> 4$. We work instead with the moment-generating function $M(t) = \E[X^t]$ for normalized eigenvalues $X = λ/\AM$. Since $M'(0) = \E[\log X]$, the log-determinant becomes $\log\det(A) = n(\log \AM + M'(0))$ -- the problem reduces to estimating a derivative at $t = 0$. Trace powers give $M(k)$ at positive integers, but interpolating $M(t)$ directly is ill-conditioned due to exponential growth. The transform $K(t) = \log M(t)$ compresses this range. Normalization by $\AM$ ensures $K(0) = K(1) = 0$. With these anchors fixed, we interpolate $K$ through $m+1$ consecutive integers and differentiate to estimate $K'(0)$. However, this local interpolation cannot capture arbitrary spectral features. We prove a fundamental limit: no continuous estimator using finitely many positive moments can be uniformly accurate over unbounded conditioning. Positive moments downweight the spectral tail; $K'(0) = \E[\log X]$ is tail-sensitive. This motivates guaranteed bounds. From the same traces we derive upper bounds on $(\det A)^{1/n}$. Given a spectral floor $r \leq λ_{\min}$, we obtain moment-constrained lower bounds, yielding a provable interval for $\log\det(A)$. A gap diagnostic indicates when to trust the point estimate and when to report bounds. All estimators and bounds cost $O(m)$, independent of $n$. For $m \in \{4, \ldots, 8\}$, this is effectively constant time.

PANDORA: A Parallel Dendrogram Construction Algorithm for Single Linkage Clustering on GPU

This paper presents \pandora, a novel parallel algorithm for efficiently constructing dendrograms for single-linkage hierarchical clustering, including \hdbscan. Traditional dendrogram construction methods from a minimum spanning tree (MST), such as agglomerative or divisive techniques, often fail to efficiently parallelize, especially with skewed dendrograms common in real-world data. \pandora addresses these challenges through a unique recursive tree contraction method, which simplifies the tree for initial dendrogram construction and then progressively reconstructs the complete dendrogram. This process makes \pandora asymptotically work-optimal, independent of dendrogram skewness. All steps in \pandora are fully parallel and suitable for massively threaded accelerators such as GPUs. Our implementation is written in Kokkos, providing support for both CPUs and multi-vendor GPUs (e.g., Nvidia, AMD). The multithreaded version of \pandora is 2.2$\times$ faster than the current best-multithreaded implementation, while the GPU \pandora implementation achieved 6-20$\times$ on \amdgpu and 10-37$\times$ on \nvidiagpu speed-up over multithreaded \pandora. These advancements lead to up to a 6-fold speedup for \hdbscan on GPUs over the current best, which only offload MST construction to GPUs and perform multithreaded dendrogram construction.