Siroky Jan·Apr 14Mapping AI Bias with 1.5M QuestionsFrom Elon Musk’s critiques to Bottlecap’s 1.5M question bias scan.
Siroky Jan·Feb 6SGD, RMSProp, Adam, or AdamW? A High-Stakes Benchmark on Time-Series DataPutting the industry’s top 4 algorithms to the test against 200 years of noisy solar data.
Siroky Jan·Jan 20Searching for the Path to Minimum Loss: A Deep Dive into OptimizersTracing the mathematical evolution from Stochastic Gradient Descent to the adaptive scaling of AdamW.
Siroky Jan·Jan 12Deep Diving into Advanced Loss Functions for Time SeriesThe Search for the Soul of a Time Series: Putting Specialized Loss Functions to the Test
Siroky Jan·Jan 7The Solar Grand Prix: Pitting Loss Functions Against the Sun’s VolatilityReady, steady, go! This article compares classic loss functions applied on a real world data predictor
Siroky Jan·Dec 16, 2025The Learning Rate Finder Experiment on the Sunspot DatasetA hands-on test of the “10x rule” using real-world Kaggle data
Siroky Jan·Nov 19, 2025Unlock the Universe: Three Datasets That Will Transform Your ML SkillsFrom RNNs to CNNs: Practical Techniques for Deploying ML on Real Astronomical Data.
Siroky Jan·Nov 4, 2025Three Useful Tips for Neural Network ScriptingManaging hyperparameters, visualizing training results, and saving your work.