https://flajszer.com/https://flajszer.com/about/https://flajszer.com/notebooks/https://flajszer.com/notebooks/linear-regression-with-gradient-descent/https://flajszer.com/notebooks/logistic-regression-with-gradient-descent/https://flajszer.com/slipbox/https://flajszer.com/slipbox/a-coarser-aggregate-cant-add-information-the-finer-one-lacked-but-a-denser-one-can-expose-signal-the-sparser-one-buried/https://flajszer.com/slipbox/a-functions-robustness-is-determined-by-the-component-with-the-worst-asymptotic-growth-rate/https://flajszer.com/slipbox/a-loss-function-is-a-preference-over-error-distributions/https://flajszer.com/slipbox/a-permutation-invariant-representation-guarantees-that-the-feature-vector-depends-only-on-content-and-never-on-order/https://flajszer.com/slipbox/a-quantity-that-compounds-multiplicatively-across-steps-can-be-made-additive-by-applying-the-log-function-to-it-which-lets-the-sum-based-statistical-toolbox-apply-to-it/https://flajszer.com/slipbox/a-randomized-control-group-isolates-a-causal-effect-because-regression-to-the-mean-affects-both-arms-equally-and-cancels-in-the-comparison/https://flajszer.com/slipbox/a-replication-knob-is-a-source-of-variation-under-a-fixed-experimental-design/https://flajszer.com/slipbox/a-replication-knob-must-be-both-an-arbitrary-choice-and-unrelated-to-the-target-variable/https://flajszer.com/slipbox/a-rule-of-thumb-for-performing-propositionalization-is-starting-bottom-up-by-under-aggregating/https://flajszer.com/slipbox/a-single-observation-of-extreme-value-cant-reveal-its-own-noise-to-signal-ratio-but-repeated-observations-can-by-how-far-values-pull-back-toward-the-mean/https://flajszer.com/slipbox/an-experiment-can-be-a-comparison-or-a-gate-depending-on-the-intent-not-its-mechanics/https://flajszer.com/slipbox/averaging-an-ensembles-predictions-preserves-shared-signal-and-cancels-uncorrelated-errors/https://flajszer.com/slipbox/avoiding-outlier-magnitudes-makes-the-measure-robust/https://flajszer.com/slipbox/because-the-true-value-of-a-parameter-is-unreachable-the-bayesian-expresses-it-as-a-distribution-over-possible-values/https://flajszer.com/slipbox/choosing-a-loss-means-deciding-the-price-of-errors/https://flajszer.com/slipbox/decision-trees-select-the-feature-and-its-split-threshold-of-a-node-based-on-the-sum-of-errors-of-their-child-leaves/https://flajszer.com/slipbox/distribution-shape-determines-which-measure-of-location-is-trustworthy/https://flajszer.com/slipbox/every-data-point-is-made-of-signal-and-noise/https://flajszer.com/slipbox/extreme-values-regress-toward-the-mean-because-they-carry-noise-that-doesnt-repeat/https://flajszer.com/slipbox/frequentists-put-the-probability-on-the-data-believing-the-parameter-is-fixed-while-bayesians-put-the-probability-on-the-parameter-and-believe-the-data-is-fixed/https://flajszer.com/slipbox/git-worktrees-allow-working-on-multiple-branches-in-parallel/https://flajszer.com/slipbox/group-leakage-happens-when-evaluating-generalization-but-splitting-at-the-observation-level/https://flajszer.com/slipbox/harmonic-mean-is-sensitive-to-small-values/https://flajszer.com/slipbox/log-space-target-can-be-used-to-encode-proportional-error-price-across-the-target-distribution/https://flajszer.com/slipbox/mad-rescaled-by-1-4826-is-comparable-to-under-a-gaussian-distribution/https://flajszer.com/slipbox/mean-and-median-can-be-misleading-when-the-distribution-is-multimodal/https://flajszer.com/slipbox/mean-calculated-on-a-weighted-sample-a-weighted-mean-of-the-population/https://flajszer.com/slipbox/measures-of-location-and-dispersion-are-complementary/https://flajszer.com/slipbox/measures-of-location-represent-position-and-dispersion-represents-distance/https://flajszer.com/slipbox/modeling-and-monitoring-are-inverses/https://flajszer.com/slipbox/noise-can-be-absolute-randomness-or-relative-to-some-target/https://flajszer.com/slipbox/on-standardized-axes-the-regression-slope-of-a-linear-function-tells-us-the-effect-of-regression-to-the-mean-on-the-data/https://flajszer.com/slipbox/outliers-mask-themselves-under-2-detection-because-they-inflate-the/https://flajszer.com/slipbox/over-aggregation-is-irreversible-under-aggregation-is-recoverable/https://flajszer.com/slipbox/parallel-rates-add-sequential-rates-dont/https://flajszer.com/slipbox/plain-git-makes-it-difficult-to-work-on-multiple-branches-in-parallel/https://flajszer.com/slipbox/random-splitting-temporal-data-makes-the-target-interpolation-not-forecasting/https://flajszer.com/slipbox/reciprocals-of-sequential-rates-are-additive-when-the-segment-lengths-are-equal/https://flajszer.com/slipbox/reproducibility-is-the-ceteris-paribus-condition-that-makes-replication-variance-a-clean-measurement/https://flajszer.com/slipbox/robust-outlier-detection-requires-robustness-on-both-location-and-dispersion-layers/https://flajszer.com/slipbox/run-variation-sources-the-axes-along-which-ceteris-paribus-can-be-violated/https://flajszer.com/slipbox/seed-granularity-shared-vs-per-source-is-a-design-choice-between-measuring-total-noise-and-attributing-it-to-sources/https://flajszer.com/slipbox/target-leakage-happens-when-training-set-contains-a-feature-carrying-target-information-unavailable-at-inference-time/https://flajszer.com/slipbox/test-sets-split-structure-determines-the-claim-about-the-model/https://flajszer.com/slipbox/the-2-outlier-rule-assumes-a-gaussian-distribution/https://flajszer.com/slipbox/the-model-will-sacrifice-errors-the-loss-considers-cheap-in-order-to-minimize-the-errors-the-loss-considers-expensive/https://flajszer.com/slipbox/the-split-type-has-to-preserve-the-structure-youre-generalizing-along/https://flajszer.com/slipbox/the-split-unit-follows-what-you-generalize-to/https://flajszer.com/slipbox/the-stage-in-workflow-components-form-a-handoff-chain/https://flajszer.com/slipbox/the-stage-in-workflow-dimension-classifies-experiments-by-which-artifact-is-the-subject-of-iteration-at-this-stage/https://flajszer.com/slipbox/trimmed-mean-is-the-dial-between-a-mean-and-the-median/https://flajszer.com/slipbox/unlike-inter-hypothesis-running-intra-hypothesis-experiments-allows-comparing-models-ceteris-paribus/https://flajszer.com/slipbox/using-mutual-information-first-allows-identifying-which-features-carry-any-signal-regardless-of-their-relationship-type/https://flajszer.com/slipbox/validation-and-test-sets-must-have-identical-split-structure-to-support-the-same-claim/https://flajszer.com/slipbox/validation-sets-split-structure-selects-the-model/https://flajszer.com/slipbox/weighted-mean-allows-introducing-encoded-preference-to-the-measure/https://flajszer.com/slipbox/whats-noise-and-whats-signal-in-a-data-point-depends-on-how-the-target-variable-is-defined/https://flajszer.com/slipbox/zettelkasten-notes-should-be-claim-based/https://flajszer.com/slipbox/zettelkastens-value-comes-from-complex-conclusions-emerging-from-atomic-claims/https://flajszer.com/work/