Series
Machine Learning Series 1
14 parts, in reading order. Each one builds on the last.
- 01
Machine learning for quantitative finance in ten ideas
AUG 31 2026 · 6 min
Ten ideas carry machine learning for finance: the target decides the problem, the data is part of the model, a linear model is a weighted sum, every feature has…
- 02
What machine learning is: rules learned from examples
SEP 1 2026 · 12 min
Machine learning is teaching a computer by example instead of by instruction: you show it many past cases with the right answers, it works out the pattern on…
- 03
The data is part of the model: splits, gaps and outliers
SEP 3 2026 · 14 min
Which rows a model trains on, how the test rows are chosen, what is done about missing values, and how far one wild value can pull a fit all change the result…
- 04
Regression vs classification: predicting a number or a label
SEP 8 2026 · 9 min
Most machine learning answers one of two questions: "how much?" (regression, the answer is a number) or "which one?" (classification, the answer is a category).
- 05
Linear regression: the first model, a straight line
SEP 14 2026 · 13 min
Linear regression draws the straight line that best fits your data, where "best" means the total of the squared misses is as small as possible; each number it…
- 06
Every feature has a timestamp: leakage, scaling and twin features
SEP 16 2026 · 10 min
For every feature a model uses, ask when its value became knowable and whether you could compute it live.
- 07
Decision trees: a model made of yes/no questions
SEP 22 2026 · 12 min
A decision tree makes a prediction by asking a short series of yes/no questions about the input, like a flowchart an expert would draw; it learns which…
- 08
Nearest neighbours: predict what similar past cases did
SEP 26 2026 · 10 min
k-nearest neighbours keeps every training row and, asked about a new case, finds the k most similar old ones and averages their answers.
- 09
Bias and variance: where a model's errors come from
SEP 30 2026 · 11 min
A model's error on new data comes from three places: bias (it is too simple to capture the real pattern), variance (it changes too much depending on which data…
- 10
Scoring a model: a number means nothing without a split, a metric and a baseline
OCT 1 2026 · 10 min
A model's score is meaningless on its own.
- 11
Bagging: many copies of a model, averaged
OCT 6 2026 · 11 min
Bagging trains many copies of the same model, each on a slightly different random resample of the data, and averages their answers; the copies make different…
- 12
Random forests: bagging with one extra rule
OCT 12 2026 · 10 min
A random forest is bagging with one extra rule: at every split, each tree may only choose from a random handful of the features.
- 13
Boosting: models that learn from the last one's mistakes
OCT 14 2026 · 12 min
Boosting builds a strong model from many weak ones trained in sequence, each focusing on what the previous ones got wrong.
- 14
Choosing a model is a systems decision
OCT 15 2026 · 9 min
Accuracy is one column in the table. Training time, time per prediction and its worst case, whether the model keeps the data, whether it can be explained, how…