Machine learning cheat sheet (Computer vision)

1. 1Python for dataThe everyday toolkit: NumPy arrays and vectorised maths, pandas for tables, notebooks for exploration, charts, and the file formats data arrives in.8 concepts

   1. 1NumPy arraysN-dimensional arrays holding one data type, with a shape and a dtype, stored in contiguous memory and operated on by fast compiled code.
   2. 2Vectorisation and broadcastingVectorisation applies an operation to whole arrays at once instead of looping in Python.
   3. 3pandas DataFramesA table of labelled columns, each a typed Series, with a row index.
   4. 4groupby, merge and reshapinggroupby splits rows into groups and aggregates each one; merge joins tables on keys like SQL; pivot_table and melt switch between wide and long formats.
   5. 5Jupyter notebooksInteractive documents that mix code cells, output, charts and notes, run by a kernel that keeps variables alive between cells.
   6. 6Matplotlib chartsThe base plotting library for Python; seaborn adds statistical charts on top.
   7. 7CSV, JSON and ParquetCSV is plain-text rows with no types; JSON holds nested records, often one per line (JSON Lines); Parquet is a compressed, columnar binary format that stores…
   8. 8Plain Python for data workBefore NumPy and pandas come lists, dicts, comprehensions, functions, f-strings and reading files with with open(...).

   8 in this chapterOpen this chapter
2. 2The maths you needJust enough linear algebra, calculus, probability and statistics to understand what models compute, how they learn and whether a result is real.9 concepts

   1. 1Vectors, matrices and the dot productA vector is an ordered list of numbers, such as one row of features; a matrix is a grid, such as a dataset with n rows and d columns.
   2. 2Matrix multiplication and shapesMultiplying an (n, k) matrix by a (k, m) matrix gives an (n, m) matrix, where each entry is the dot product of a row and a column.
   3. 3Derivatives and gradientsA derivative is the rate at which a function changes as its input changes.
   4. 4Gradient descentRepeatedly moving the parameters a small step against the gradient: w = w - lr * grad.
   5. 5The chain ruleThe derivative of a function of a function is the product of their derivatives: if y = f(g(x)), then dy/dx = f'(g(x)) times g'(x).
   6. 6Probability and Bayes' theoremConditional probability P(A|B) is the chance of A given that B happened.
   7. 7Probability distributionsDescriptions of how likely each value is: normal (the bell curve, set by mean and standard deviation), Bernoulli and binomial (yes or no outcomes), Poisson…
   8. 8Descriptive statisticsNumbers that summarise data: mean and median for the centre, variance and standard deviation for spread, percentiles for shape, and correlation for the linear…
   9. 9Hypothesis tests and p-valuesA test asks how surprising the data would be if there were no real effect (the null hypothesis); the p-value is that probability.

   9 in this chapterOpen this chapter
3. Data cleaning and feature engineeringTurning raw data into model-ready features without fooling yourself: exploration, missing values, outliers, encoding, scaling, pipelines and imbalanced classes.9 conceptsOpens with credits
4. Classical machine learningThe models that still win on most tabular problems: linear and logistic regression, regularisation, trees, forests and boosting, plus clustering and PCA for data without labels.10 conceptsOpens with credits
5. Evaluating and tuning modelsGetting an honest estimate of how a model will do on new data: splits, cross-validation, leakage, the right metrics, baselines, tuning and explaining what the model relies on.9 conceptsOpens with credits
6. Neural networks and trainingHow neural networks compute and learn: layers and activations, tensors and autograd, losses, backpropagation, optimisers, the PyTorch training loop and how to read loss curves.9 conceptsOpens with credits
7. Deep learning architecturesThe architectures behind modern vision and language models: embeddings, CNNs, recurrent networks, attention and transformers, and how to adapt pretrained models to your task.7 conceptsOpens with credits
8. Computer vision in practiceWorking with images end to end: pixels as arrays, OpenCV, labelled datasets, augmentation, detection, segmentation, vision transformers, the metrics that judge them and exporting a model to run on a device.10 conceptsOpens with credits
9. Production ML and MLOpsTaking a model from notebook to a reliable service: tracking and versioning, serving, avoiding training-serving skew, CI/CD, monitoring for drift and deciding when to retrain.8 conceptsOpens with credits
10. 10Glossary, A to ZEvery concept in the sheet and the smaller terms it uses, in one list.225 terms

Data, models, deep learning.