Richard MichaelinTowards Data ScienceInfinitely Wide Neural-Networks | Neural Tangents Explaineduse Python and JAX to efficiently build infinitely wide networks for deeper network insights on your finite machine.Dec 14, 20201Dec 14, 20201
Richard MichaelinTowards Data ScienceOptimizing Hyperparameters the right WayEfficiently exploring the parameter-search through Bayesian Optimization with skopt in Python. TL;DR: my hyperparameters are always better…Sep 30, 20201Sep 30, 20201
Richard MichaelinTowards Data ScienceMillennial Suicides — an Analysis about ChangeA simple, yet meaningful Pyro model to uncover change-points over time.Aug 5, 20201Aug 5, 20201
Richard MichaelinTowards Data ScienceImprove Your Model with Missing Data | Imputation with NumPyroMake the best of missing data the Bayesian way. Improve model performance and make comparative benchmarks using Monte Carlo methods.Jul 28, 20202Jul 28, 20202
Richard MichaelinTowards Data Science3 Probabilistic Frameworks You should know | The Bayesian ToolkitBuild better Data Science workflows with probabilistic programming languages and counter the shortcomings of classical ML.Jul 20, 20201Jul 20, 20201
Richard MichaelinTowards Data ScienceSingle-Parameter Models | Pyro vs. STANModeling U.S. cancer-death rates with two Bayesian approaches: MCMC in STAN and SVI in Pyro.Jul 15, 2020Jul 15, 2020
Richard MichaelinTowards Data Science10.000 hours in Data Science | Gaining ProficiencyThe way from data novice to professional.Jun 26, 2020Jun 26, 2020
Richard MichaelinTowards Data ScienceCompute the Incomputable | How SVI and ELBO workOne reason why Bayesian Modeling works with real world data. The approximate light-house in the sea of randomness.Jun 22, 2020Jun 22, 2020
Richard MichaelinTowards Data ScienceHow your model is optimized | Know your OptimizationThe answer to: “Why is my model running forever?” or the classic: “I think it might have converged?”Jun 15, 2020Jun 15, 2020
Richard MichaelinTowards Data SciencePyro Top-Down Forecasting | Application-caseConnect the dots over time and forecast with confidence(-intervals).Jun 8, 20201Jun 8, 20201