{"@attributes":{"version":"2.0"},"channel":{"title":"Jan Bauer","link":"https:\/\/japhba.github.io\/","description":"Recent content on Jan Bauer","generator":"Hugo","language":"en-us","lastBuildDate":"Sat, 13 Jun 2026 00:00:00 +0000","item":[{"title":"ROC & AUC, animated","link":"https:\/\/japhba.github.io\/misc\/roc-auc\/","pubDate":"Sat, 13 Jun 2026 00:00:00 +0000","guid":"https:\/\/japhba.github.io\/misc\/roc-auc\/","description":"An interactive walk-through of the ROC curve and the area under it (AUC), built around Maneesh Sahani&rsquo;s signal-detection slides \u2014 two response distributions P(n|s\u2080) and P(n|s\u2081), a sliding criterion, and the hit-rate (TPR) \/ false-alarm-rate (FPR) trade-off it traces out. Drag the criterion on either plot, or hit play to sweep it; change the separation d\u2032 to watch AUC = \u03a6(d\u2032\/\u221a2) grow. Axes, confusion-matrix terminology and the no-discrimination diagonal follow the Wikipedia ROC article."},{"title":"Intelligence vs Cost, Speed & Price","link":"https:\/\/japhba.github.io\/misc\/intelligence-tradeoffs\/","pubDate":"Wed, 10 Jun 2026 00:00:00 +0000","guid":"https:\/\/japhba.github.io\/misc\/intelligence-tradeoffs\/","description":"Interactive scatter of the Artificial Analysis Intelligence Index against any one or two of cost (USD to run the whole Intelligence Index eval), median output speed, blended price ($\/1M tokens), latency, or size. Tick one axis for a 2-D plot, two for a WebGL (Three.js) 3-D plot you can orbit \u2014 with depth stems and a translucent Pareto-frontier surface over the non-dominated models. Live data via a cached server-side mirror, with a baked-in snapshot fallback."},{"title":"Culinary choices","link":"https:\/\/japhba.github.io\/post\/google_maps\/","pubDate":"Sat, 10 Feb 2024 17:00:00 +0000","guid":"https:\/\/japhba.github.io\/post\/google_maps\/","description":"<p>Suppose you are in a new city and looking to get dinner with friends. You pull out Google Maps to help with the decision. To make the process easier, you decide to look at the ratings: A nearby Cambodian restaurant boasts 4.8 stars, but the equally close Italian place is a close competitor at 4.6 stars, but has five times the number of reviews. Surely that must make a difference? You search for your scratchpad, promising to your friends that you got the situation.<\/p>"},{"title":"Can random actions be optimal?","link":"https:\/\/japhba.github.io\/post\/stochastic_policy\/","pubDate":"Sat, 27 Jan 2024 00:00:00 +0000","guid":"https:\/\/japhba.github.io\/post\/stochastic_policy\/","description":"<p>Is random behavior helpful in any situation? By definition, random actions are the most uninformed, and if any better is known should be suboptimal. Yet, the issue is more subtle. Reinforcement learning and game theory can be paradigms to reason about this.<\/p>"},{"title":"Do auto-regressive models bite their own tail?","link":"https:\/\/japhba.github.io\/post\/autoregressive\/","pubDate":"Wed, 27 Dec 2023 00:00:00 +0000","guid":"https:\/\/japhba.github.io\/post\/autoregressive\/","description":"<p>Autoregressive models use their output to arrive at predictions. In machine learning, this amounts to &ldquo;training on the output&rdquo;, i.e., generated data. More broadly, intelligent behavior is often accompanied by deep thought or even dreaming between actions. In both of these cases, the system is decoupled from the ground truth. Despite this apparent conundrum, there seems to be a benefit.<\/p>"},{"title":"Experience","link":"https:\/\/japhba.github.io\/experience\/","pubDate":"Tue, 24 Oct 2023 00:00:00 +0000","guid":"https:\/\/japhba.github.io\/experience\/","description":{}},{"title":"What is meta in meta-learning?","link":"https:\/\/japhba.github.io\/post\/learning\/","pubDate":"Sat, 03 Jun 2023 00:00:00 +0000","guid":"https:\/\/japhba.github.io\/post\/learning\/","description":"Meta-learning summarizes the concept of learning a more general framework to learn \u2013 hence the name. Yet, this concept subsumizes a range of multiple concepts, including transfer learning, few-shot learning, continual learning, and fine-tuning. We develop an abstracted framework that unifies these notions. This extends beyond parametric models."},{"title":"Mean-field decoupling via auxiliary variables","link":"https:\/\/japhba.github.io\/post\/dmft\/","pubDate":"Thu, 10 Nov 2022 22:00:00 +0000","guid":"https:\/\/japhba.github.io\/post\/dmft\/","description":"<p>In statistical physics, we are often dealing with systems that comprise many components. In order to calculate their statistics, high-dimensional integrals over those variables $\\boldsymbol{x}\\in\\mathbb{R}^{N}$ with $N\\gg1$ are required. A typical form is<\/p>"},{"title":"Example Talk","link":"https:\/\/japhba.github.io\/event\/example\/","pubDate":"Sun, 01 Jan 2017 00:00:00 +0000","guid":"https:\/\/japhba.github.io\/event\/example\/","description":"An example talk using Hugo Blox Builder&rsquo;s Markdown slides feature."}]}}