Machine Learning
Cross-validation for groups and time: check what a split holds out
Audit random, grouped and chronological folds on the same small table, and choose a split that matches the prediction you intend to make.
tutorial4 min read
Articles
Practical guides to learning AI and machine learning, written by Gaurav Jain. Worked examples, projects with complete code and learning plans you can follow.
Page 2 of 8
- Machine Learning
Decision-tree split criteria: calculate weighted Gini impurity
Enumerate thresholds on an explicit binary dataset, fit a one-level tree and compare held-out performance with a majority baseline.
tutorial4 min read
- Maths
Derivatives for machine learning: check a squared-error gradient
Derive the weight and bias gradients of a linear model, verify dimensions and compare central differences with the analytic result.
tutorial4 min read
- Generative AI
Diffusion models step by step with a one-dimensional Gaussian
Calculate the forward noising distribution and an exact reverse conditional on a declared Gaussian, separating an analytic oracle from a learned denoiser.
tutorial4 min read
- Statistics
Entropy, cross entropy and KL divergence on the same distribution
Calculate uncertainty, coding mismatch and KL divergence in nats, verify their identity and inspect a zero-probability failure.
tutorial4 min read
- Maths
Gradient descent learning rates: convergence, oscillation and divergence
Run three learning rates on an explicit quadratic, derive the stability boundary and compare reproducible objective curves.
tutorial4 min read
- Generative AI
LLM tool calling with validated arguments and a bounded dispatcher
Treat tool proposals as untrusted input, validate a read-only unit conversion and reject unknown tools, fields and nonfinite values.
tutorial4 min read
- Maths
Log-sum-exp: calculate model scores without exponential overflow
Derive the maximum-shift identity, reproduce overflow and verify stable log-sum-exp and softmax calculations in Python.
tutorial4 min read
- Machine Learning
Logistic regression from scratch: fit probabilities and inspect errors
Implement stable binary cross entropy and its gradient in NumPy, train on invented data and compare held-out predictions with a constant baseline.
tutorial4 min read
Learn the subject in order
The courses take each idea from the intuition and the maths to code you write yourself, from your first line of Python to machine learning.