The Incerto books are written in a provocative style that may not appeal to everyone. Beneath the prose, each book explores a profound concept in probability theory. I tried to strip each to its bare bones. (1/6)
I think of Taleb's viewpoints as "probabilistic stoicism": accept the randomness and unpredictability in the world, build a system and accept consequences to maximize expected success, however you define it. (6/6)
Skin in the Game emphasizes the necessity of personal exposure to the consequences of risk. By facing an 'absorbing barrier' (potential ruin), individuals are forced into more robust behavior, allowing the system to eliminate the fragile over time and improve. (5/6)
Antifragile explores how systems respond to randomness. An antifragile entity exhibits a convex response, meaning it benefits from higher volatility (Jensen's Inequality). Example: buffers, good selection criteria.... (4/6)
The Black Swan highlights rare yet extremely impactful events. It argues the world isn't governed by 'Mediocristan' (compact Gaussian distributions), but by 'Extremistan' with fat tails – assuming we can even characterize the distribution. (3/6)
Fooled by Randomness states that a string of successes is likely due to chance, especially when some selection is involved. Randomness produces patterns, and our brains are wired to detect those. (2/6)
The Incerto books are written in a provocative style that may not appeal to everyone. Beneath the prose, each book explores a profound concept in probability theory. I tried to strip each to its bare bones. (1/6)
The KERMIT (kermit.ugent.be) research unit at @fbwugent.bsky.social focuses on holistic modeling for life sciences. In my subunit, we are particularly interested in (stochastic, differentiable, evolutionary...) computation for plants and biotechnology.
You can (an many people obviously did) go much farther than my example. I have been thinking about Von Neumann boundary conditions (CST fluxes), though not entirely clear.
As everyone shares AI-generated images, I'll also post something I've created. I took up watercolors to relax, and my latest one I am quite happy about!
This beautiful paper extends the bias-variance trade-off for some losses to a bias-variance-diversity trade-off for ensembles. Diversity of the predictors always reduces the risk but might lead to more variance or bias.
Low-key, I'm a bit depressed about having lost the Signal chat history with a new phone. I still have the history on the Desktop app, but I am afraid of relinking and losing years of memories and photos I shared with my partner.
Proud to have finally finished the new syllabus of my modeling course. 😌 We cover the fundamentals of modeling with ODEs and probability. I hope the students will also like it in a couple of weeks.
