Statystyka
What does a typical value hide?
Change one record and discover when the mean and median tell different stories.
How much spread does one extreme value create?
Compare standard deviation, quartiles, and histograms while changing one observation.
Mean and variance: where are the values, and how spread out are they?
Calculate mean, deviations, variance, and standard deviation; discover how shifting and scaling change them.
How do you sample from a probability distribution?
Turn probabilities into outcomes with uniform draws, cumulative probabilities, and inverse transform sampling.
What does a random seed reproduce in PyTorch?
Trace pseudorandom generator state, reset and resume PyTorch generators, and separate seeds from sampling probabilities.
Does filling missing values restore the data?
Investigate how missingness and imputation change a column without recovering its hidden values.
Are twelve records twelve independent examples?
See why repeated entities can make a validation score answer the wrong question.
Can zero correlation hide a predictable pattern?
Compare a straight line, a curve, and a constant before interpreting Pearson correlation.
Which rows are allowed to fit a scaler?
Track how fitting preprocessing on validation data changes the training representation.
Can 95% accuracy mean complete failure?
Keep 200 answer records fixed and discover what accuracy, precision, and recall reveal about a failure detector.
Why does attention divide dot products by √d?
Discover why attention divides dot products by √d, and distinguish variance scaling, cosine similarity, and softmax.