Accurate and fast classification losses for SDCA and neural networks: cross-entropy, top-k, multiclass and multilabel SVM, and distillation losses for PyTorch, JAX, NumPy, Rust, C/C++ and MATLAB
-
Updated
Nov 16, 2026 - Python
Accurate and fast classification losses for SDCA and neural networks: cross-entropy, top-k, multiclass and multilabel SVM, and distillation losses for PyTorch, JAX, NumPy, Rust, C/C++ and MATLAB
Leveraging network motifs to improve artificial neural networks
Bilimsel Hesaplama ve Sayısal Analiz masterclass'ı. Hazır kütüphaneler yok; saf matematik ve NumPy var. Makine epsilonu analizi, güvenli hibrit kök bulma, pivotlu LU/Cholesky çözücüler, Chebyshev interpolasyonu, uyarlamalı Gauss kuadratürü ve katı (stiff) diferansiyel denklemlerin (ODE/BVP) sıfırdan inşası.
➗ This work presents a thorough analysis of a quadratic optimization model, confirming convexity, coercivity, and the exact global minimum. It compares gradient descent and Newton’s method, highlighting Newton’s superior efficiency when the Hessian is invertible.
Validate PyTorch model outputs across NVIDIA, AMD, Trainium, TPU, and Apple Silicon. Find numerical divergence before it reaches production.
High-performance, branchless numerical stability kernels and compiler-optimized core infrastructure for advanced JAX/XLA deep learning architectures.
Softmax Collapse-free log_softmax, softmax and cross_entropy for PyTorch, JAX and TensorFlow (pip install softmax1p)
Research project analyzing stability and robustness of deep learning optimizers (SGD, Adam, SAM) under label noise and precision constraints.
Certify that your top-k, argmin or threshold was determined by your data and not by the rounding of the kernel it ran on — or refuse, and name the boundary.
A research-grade course in Numerical Methods and Optimization for ML. No deep learning libraries—just pure math and NumPy. Rebuilds automatic differentiation engines, stable SVD/Cholesky algorithms, quasi-Newton solvers (BFGS), stable log-sum-exp layers, and Gaussian Process regressions from scratch
Detect numerical instability in ML applications using learned invariants (Soft Assertions) without modifying model logic. ACM FSE 2025.
Pre-registered experiment on differentiable EML-tree symbolic regression (Odrzywolek 2026, arXiv:2603.21852). Streamlit UI + headless reproducer + full results.
R package and replication code for the article “Numerical stability enhancements in beta autoregressive moving average model estimation” by Cribari-Neto, F., Costa, E., and Fonseca, R. V., published in the Brazilian Journal of Probability and Statistics (2025). DOI: 10.1214/25-BJPS645.
量測 LLM 推論的決定性:同一 prompt 重複執行時,輸出從第幾個 token 開始分歧。
Interactive Python Softmax simulator with step-by-step probability normalization, temperature scaling, numerical stability comparison, entropy, cross-entropy, and simplex visualization.
Reproducible audit of finite-range failures in ONNX Normalizer implementations
Classical vs. modified Gram-Schmidt: C and MATLAB implementations, with error analysis on ill-conditioned matrices. Orthogonalisation de Schmidt classique vs. modifiée : implémentations C et MATLAB, analyse d'erreur sur matrices mal conditionnées.
Incremental corpus centroid with exact add/remove: compensated float64 accumulation and drift stats that separate rotation from diversity
Makine Öğrenmesi için Lineer Cebir ve Matris Hesaplamaları masterclass'ı. Hazır kütüphaneler yok; saf geometrik sezgi ve NumPy var. Sıfırdan Gauss eleme, LU ve QR ayrıştırmaları, kararlı en küçük kareler, özdeğer iterasyonları ve doruk noktası olarak PCA ile uçtan uca SVD tabanlı boyut indirgeme mimarisi.
Symbolic generation, Fortran codegen and nonlinear search of stable high-order summation-by-parts boundary closures
To associate your repository with the numerical-stability topic, visit your repo's landing page and select "manage topics."