Skill UI

Probabilistic deep-learning framework built on PyTorch for single-cell omics; handles batch correction, integration, multimodal and spatial data, differential expression, and transfer learning with unified APIs.

This skill provides zero-shot time series forecasting using Google's TimesFM foundation model. It forecasts any univariate time series (like sales, sensor data, or weather) without requiring custom model training. The model delivers point forecasts along with calibrated prediction intervals (quantiles), offering a robust and probabilistic view of future trends. It is ideal for rapid, reliable forecasting across diverse time-based datasets.

PyMC is a powerful Python library for probabilistic programming and advanced Bayesian inference. This comprehensive guide details the full workflow for building, fitting, and validating complex Bayesian models. Learn to implement hierarchical structures, perform MCMC sampling (NUTS), conduct posterior predictive checks (PPC), and effectively quantify uncertainty using industry best practices.

PyMC is a powerful Python library designed for probabilistic programming and sophisticated Bayesian modeling. It enables users to build complex models, such as hierarchical structures and time series, and perform rigorous inference using methods like MCMC (NUTS) sampling and Variational Inference. Key features include model comparison (LOO, WAIC) and comprehensive diagnostic tools for uncertainty quantification.

scvi-tools is a comprehensive Python framework utilizing deep generative models and variational inference for analyzing diverse single-cell data modalities. It excels in complex analyses such as probabilistic batch correction (scVI), multi-modal integration (MultiVI), and differential expression analysis with uncertainty. It supports single-cell RNA-seq, ATAC-seq, and various multi-omics datasets, providing advanced tools for modern genomic research.

Utilizes Google's TimesFM foundation model for zero-shot time series forecasting. It processes any univariate time series (e.g., sales, sensor data, weather) without requiring custom model training. The skill provides probabilistic forecasts with calibrated prediction intervals (quantiles) and includes a mandatory system checker to ensure sufficient RAM, GPU, and disk resources for safe local inference.

A specialized skill for achieving asymptotic performance gains when classical algorithms are already near optimal. It leverages advanced mathematical techniques and probabilistic data structures (e.g., Bloom Filters, HyperLogLog, FFT, MinHash/LSH) to handle massive datasets (n ≥ 10⁶) by accepting bounded approximations or exploiting deep structural properties. Requires strict adherence to error parameter and complexity justification protocols.

This guide provides rigorous methodology for authors submitting security and machine learning research to top conferences. It emphasizes the need to 'freeze the world' of measurement by pinning target versions, microarchitectural states, and network conditions. It details how to report probabilistic evidence (like attack success rates) and manage ethical data release, ensuring findings remain valid and verifiable over time.

This guide details the methodology for packaging scientific artifacts (code, solvers, data, logs) for machine learning submissions. It focuses on creating double-blind, self-contained ZIP archives that allow reviewers to independently and quickly verify complex probabilistic claims, even under strict time constraints. It covers best practices for anonymity, structured README generation, and handling data/licensing.

This guide details advanced best practices for achieving scientific reproducibility in probabilistic modeling and deep learning research. It mandates rigorous disclosure of seeds, hyperparameters, diagnostic traces (e.g., R-hat, ELBO), and compute environments to ensure that all claims made in academic papers are verifiably backed by traceable evidence. It elevates reproducibility from a simple code run to a comprehensive audit.

This guide helps researchers determine the most appropriate conference (UAI, AISTATS, NeurIPS, ICML, etc.) for their work, particularly when the core contribution revolves around uncertainty representation, probabilistic reasoning, or causal inference. It provides detailed filters and comparison points to refine the paper's scope and target audience.

A comprehensive guide for authors submitting to venues focused on probabilistic AI, including UAI. It emphasizes achieving maximum formal rigor by enforcing strict notational discipline, clearly separating conditioning from intervention, explicitly stating assumptions, structuring the main body for maximum impact, and calibrating claims to match empirical evidence. Essential for papers involving graphical models, causality, and Bayesian methods.