• A Whole New Dimension of Optimization

    Numerical Optimization: Optimizing multivariate objectives.

  • Take the Rough With the Smooth

    Numerical Optimization: How helpful are smooth objectives? This post explores strategies for incorporating derivatives in optimization algorithms.

  • The Roots of No Evil

    The first post in a planned series of posts about numerical optimization algorithms. This one is about extrema and roots.

  • Random Integers

    Efficient sampling from a discrete distribution is a useful yet nontrivial algorithmic building-block, which involves some interesting and clever ideas.

  • Machine Learning. Literally.

    Arguably, the most successful application of machine learning is largely unknown to most practitioners. Appropriately, it literally involves machines that learn.

  • Laws, Sausages and ConvNets

    The nuts and bolts of Convolutional Neural Networks: algorithms, implementations and optimizations.

  • The Generative-Discriminative Fallacy

    Machine learning algorithms are often categorized as either discriminative or generative. While this dichotomy can be instructive, it is often misleading.

  • Learning Dynamical Systems

    Machine Learning meets Differential Equations.

  • The Name of The Rose

    Convergent evolution is a common phenomenon in machine learning: many dissimilar scenarios lead to similar algorithms. When it comes to generalizations, though, distinctive underlying ideas could be fundamental.

  • Pointless Topology via Abstract Nonsense

    Trigger Warning: pure mathematics. Stone Duality gives a rigorous meaning to the slogan “Geometry is dual to Algebra”.

  • Meta-Sequences in C++

    With the introduction of variadic templates to C++, meta-sequences became a central idiom in meta-programming. The standard implementation is not always the best choice.

  • Random Bitstreams

    Controlling the entropy of pseudo-random bits in Python when performance matters.

subscribe via RSS