Bayesian priors and Ridge regression
Showing the equivalence of Bayesian priors and Ridge regression
Choose your optimization algorithm carefully
How does the choice of optimization result impact how long time it takes to find the correct solution
Confidence intervals - Not what you think
Confidence intervals actually says nothing at all about the probability of the value being in the interval
Deep Evidential Modeling
Deep Evidential Modeling for safe, robust and accurate machine learning applications
Identifiability and granularity for time series models
Is it better to aggregate sparse data or does the granularity add value
Impact of bimodality in regression problems
A few thoughts on apparent bimodality for regression problems!
Borch: A Deep Universal Probabilistic Programming Language
Borch: a scalable deep universal probabilistic programming language, built on top of PyTorch.
The apparent success of the maximum likelihood principle
The maximum likelihood principle is modeling without assumptions, and there is no such thing. The assumption of no assumption is a flawed assumption.
The importance of context
The importance of context when building models
The insignificance of significance
Statistical significance has always held a slightly magical status in the research community, this is misguided ans should be stopped
The truth about priors and overfiting
Learn about the effect of week and strong priors and how it effects overfitting
What does the prior actually do in a Bayesian analysis?
How the prior and posterior effects the likelihood in Bayesian inference, and why setting the correct priors are important.
On the fallacy of A/B testing
A/B testing is a commonly used method for determining the effectiveness of different variations of a product or marketing campaign. However, despite its widespread use, A/B testing has several limitations that make it a flawed methodology, particularly when compared to bespoke machine learning models.
Ready for the next step?
Tell us the decision you need to make with confidence. We'll tell you honestly whether a model can help.