ML without tears

Machine Learning without tears

Mathy stuff, how I would have liked to learn them

  • Retirement, Stopping times and Bandits: The Gittins index

    “A colleague of high repute asked an equally well-known colleague:— What would you say if you were told that the multi-armed bandit problem had been solved?— Sir, the multi-armed bandit problem is not of such a nature that it can be solved.” Peter Whittle In our busy daily life, while multi-tasking we are constantly faced…

  • Make any predictor uncertainty-aware via conformal prediction

    The importance of being uncertainty-aware. When making a prediction (in a regression or a classification setting) based on observed inputs , it is often important to know which range  the true value will fall in, with high confidence. For instance, a trader would be interested in knowing within which boundaries the stock price remains, rather than a…

  • The geometry of Gaussian variables

    Gaussian (or normal) variables are all around the place. Their expressive power is certified by the Central Limit Theorem, stating that the mean of independent (and not necessarily Gaussian!) random variables tends to a Gaussian variable. And even when a variable is definitely not Gaussian, it is sometimes convenient to approximate it as one, via Laplace…

  • Is Reinforcement Learning all you need?

    When attacking a new problem, the algorithm designer typically follows 3 main steps: When reporting her/his work, the algorithm designer will proudly focus on step 3), briefly mention 2) and likely sweep 1) under the carpet. Yet, skimming alternatives off is a crucial step, that inevitably impacts (positively or negatively) months of hard work on…

  • Policy gradient for black-box optimization

    Policy gradient method are widely used in the Reinforcement Learning settings. In this post we build policy gradient from the ground up, starting from the easier static scenario first, where we maximize a reward function depending solely on our control variable . In subsequent posts, we will turn our attention to the contextual bandit setting,…

Get In Touch