crps

Cryptanalysis of Physically Unclonable Functions

Ensemble price forecasting with volatility prediction (XGBoost on cached features). Multiple simulated paths per request; CRPS scoring for calibration and sharpness. Synthetic price data for options and portfolio analytics. Python, XGBoost, NumPy, Pandas, properscoring, Pyth API, PostgreSQL, Pydantic, Docker.

Official pytorch implementation codes for NeurIPS-2023 accepted paper "Distributional Learning of Variational AutoEncoder: Application to Synthetic Data Generation"

PyTorch-based implementations of the Continuously-Ranked Probability Score (CRPS) as well as its locally scale-invariant version (SCRPS)

Model stacking for predictive ensembles

DINOv3 as a Frozen Encoder for CRPS-Oriented Probabilistic Rainfall Nowcasting (Neurips W4C Competition)

Official implementation of 'Distributional Learning of Variational AutoEncoder: Application to Synthetic Data Generation' (DistVAE) with pytorch (NeurIPS 2023 accepted paper).

Distributional crypto-return forecasting via Wasserstein-geodesic extrapolation in quantile-function space. WGeo family wins 12/12 (asset × horizon) cells over 6.75y walk-forward CRPS vs GARCH and classical baselines. v0.4.

Implementation of"Learning Quantile Functions for Temporal Point Processes with Recurrent Neural Splines" (@ AISTATS 2022)

Reproducibility artifact for 'Heads, Not Backbones: Output Heads Dominate Architectures on Fat-Tailed Returns' (arXiv:2606.30037)

Quantified pub quiz

A modern, privacy-focused companion app designed to help people with Complex Regional Pain Syndrome (CRPS) document symptoms, identify patterns, and better understand their daily condition.

Distributional forecasts scored with CRPS and PIT calibration, plus block-bootstrap tests that separate leaderboard skill from luck

Links calibrated deep probabilistic forecasting to reinforcement-learning monetary policy through a shared belief state. A CRPS-trained multi-scale LSTM supplies time-varying uncertainty to a POMDP central bank; realistic beliefs collapse the learned Taylor coefficient from ~1.1 to ~0.02 across every seed under both PPO and SAC.

R2 — LLM calibration benchmark: 18-scenario Florida rainfall eval crossing a specificity gradient with a seasonal gradient, CRPS scoring, and ground-truth-free self-consistency checks.

Training-free probabilistic time-series forecasting with Conformal Seasonal Pools (CSP): a numpy-only empirical sampler with quantiles, prediction intervals, and rolling-origin backtesting.

An end-to-end probabilistic machine learning pipeline for retail demand forecasting. Uses LightGBM, XGBoost, and GLMs to predict full probability distributions of daily sales, evaluated via CRPS.

Repository for CRPS 2 project. Here you will find the files and firmware for multiple exoskeletons for assisting the hand of a patient

https://mp.weixin.qq.com/s/mIQYwVnpvDrxqI8Jfpxviw