AutoML Seminars Courses

GRAF - Performance Prediction with Neural Graph Features

Discover how neural graph features can predict network performance through GRAF - a topological approach that matches zero-cost proxies while offering better interpretability and consistency across tasks.

• YouTube
• 39 minutes
• Self-Paced
• Free Video

Stop Guessing, Start Discovering Trade-offs in Your ML Models

Discover how to optimize ML models beyond accuracy by using Bayesian optimization to balance trade-offs in privacy, fairness, and energy efficiency for real-world deployment.

• YouTube
• 41 minutes
• Self-Paced
• Free Video

Chronos-2 - From Univariate to Universal Forecasting

Discover Chronos-2, a pretrained model for zero-shot univariate, multivariate, and covariate-informed forecasting using group attention and in-context learning across diverse time series.

• YouTube
• 48 minutes
• Self-Paced
• Free Video

Accelerating Bayesian Inference and Data Acquisition via Amortization

Discover how amortized neural networks accelerate Bayesian inference and experimental design, enabling real-time data acquisition and reasoning under uncertainty.

• YouTube
• 41 minutes
• Self-Paced
• Free Video

Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach

Explore a novel language model architecture that scales test-time computation through latent space reasoning using a recurrent block approach, offering improved performance on reasoning benchmarks without specialized training data.

• YouTube
• 42 minutes
• Self-Paced
• Free Video

AutoML in the Age of Structured Foundation Models

Explore how structured foundation models like TabPFNv2 and Chronos are transforming AutoML, and why AutoML systems remain crucial in this new era of foundation model-driven data analysis.

• YouTube
• 43 minutes
• Self-Paced
• Free Video

Do-PFN - In-Context Learning for Causal Effect Estimation

Explore Do-PFN's innovative approach to causal effect estimation using in-context learning, eliminating the need for interventional data or known causal graphs.

• YouTube
• 46 minutes
• Self-Paced
• Free Video

Multi-Objective AutoML - Towards Accurate and Robust Models

Discover how AutoML can optimize machine learning models for both accuracy and robustness against domain shifts and input perturbations in trustworthy AI systems.

• YouTube
• 47 minutes
• Self-Paced
• Free Video

Benchmarking Beyond Statistics - Data-Driven Footprints for Explainable Black-Box Optimization

Explore data-driven benchmarking methods and algorithmic footprints to make black-box optimization transparent, reproducible, and explainable through meta-learning approaches.

• YouTube
• 31 minutes
• Self-Paced
• Free Video

ShinkaEvolve - Towards Open-Ended and Sample-Efficient Program Evolution

Discover how ShinkaEvolve leverages LLMs for efficient program evolution, achieving breakthrough results in scientific discovery with novel sampling techniques and bandit-based ensemble strategies.

• YouTube
• 47 minutes
• Self-Paced
• Free Video

Accurate Predictions on Small Data and Time Series with the Tabular Foundation Model TabPFN

Discover how TabPFN revolutionizes tabular data prediction with superior accuracy and speed, outperforming traditional methods across scientific fields while enabling fine-tuning and data generation capabilities.

• YouTube
• 50 minutes
• Self-Paced
• Free Video

Reshuffling Resampling Splits Can Improve Generalization of Hyperparameter Optimization

Discover how reshuffling resampling splits during hyperparameter optimization can enhance machine learning model performance and improve generalization on unseen data through theoretical and practical insights.

• YouTube
• 46 minutes
• Self-Paced
• Free Video

TabArena - A Living Benchmark for Machine Learning on Tabular Data

Discover TabArena, a continuously maintained benchmarking system for tabular machine learning that compares deep learning, gradient-boosted trees, and foundation models across datasets.

• YouTube
• 46 minutes
• Self-Paced
• Free Video

Tuning LLM Judge Design Decisions for 1/1000 of the Cost

Discover how to systematically tune LLM judge hyperparameters using multi-objective multi-fidelity optimization to achieve better accuracy at 1/1000th the cost while ensuring reproducibility.

• YouTube
• 42 minutes
• Self-Paced
• Free Video

Machine Learning Without Data - Training Small Models with 10 Examples

Discover how to train custom Small Language Models locally using AI-generated data instead of manual labeling, keeping sensitive information private while achieving production-ready results.

• YouTube
• 41 minutes
• Self-Paced
• Free Video