codewithasher

Vectors, matrices, the dot product, and eigenvectors — the language of machine learning.

Derivatives, the chain rule, partial derivatives, and gradients — the math behind learning.

Distributions, expectation, variance, and Bayes’ theorem — reasoning under uncertainty.

Types of ML, problem framing, train/test splits, data leakage, EDA, and preprocessing.

Precision, recall, F1, ROC/AUC, cross-validation, and the bias-variance tradeoff.

Encoding, selection, extraction, TF-IDF, time features, and handling imbalance.

Cost functions, the update rule, and learning rates — how models actually learn.

Adam, learning-rate schedules, regularization, MLE/MAP, and generalization theory.

Fit a line, measure error with MSE, and predict continuous numbers from features.

The sigmoid, decision boundaries, cross-entropy, and softmax — classifying with probabilities.

Distance metrics, majority voting, and the curse of dimensionality.

Bayes’ rule, the naive independence assumption, and a spam filter built from scratch.

Splits, entropy, information gain, and random forests — learning with yes/no questions.

Margins, support vectors, soft margins, and the kernel trick.

Bagging, gradient boosting, and XGBoost / LightGBM / CatBoost — why many models beat one.

Centroids, assignment and update steps, and choosing k — grouping unlabeled data.

Variance, covariance, and principal components — compressing data without losing signal.

DBSCAN, GMMs and EM, SVD, t-SNE, UMAP, anomaly detection, and association rules.

Bayesian networks, HMMs, MCMC, and variational inference — modeling hidden structure.

Activations, initialization, vanishing gradients, dropout, and batch/layer norm.

Convolution, filters, feature maps, and pooling — how machines see images.

Hidden state, recurrence, vanishing gradients, and gates — networks with memory.

Autoencoders, VAEs, and diffusion models — compressing and generating data.

Generator vs discriminator, the adversarial game, and mode collapse.

Tokenization, self-attention, multi-head attention, and the full transformer block.

Tokenization, embeddings, attention, and generation — build an LLM from the ground up.

Pretraining, instruction tuning, RLHF, LoRA, RAG, and serving — building real assistants.

Agents, rewards, value functions, and Q-learning — learning from rewards.

MDPs, policy gradients, actor-critic, deep RL, bandits, and offline/multi-agent RL.

Adversarial examples, causal inference, fairness, meta-learning, and graph neural nets.

Bounding boxes, IoU, the YOLO grid, anchors, and non-max suppression.

Sensors, perception, sensor fusion, planning, and control — how autonomous cars drive.

Pipelines, deployment, monitoring, drift, A/B testing, and governance.