AI From Scratch in C++

Pure C++ implementations of popular AI and machine learning algorithms without using any third-party libraries. No Python, no frameworks, no third-party libraries — just clean, idiomatic C++.

---

Weekly Schedule

Week	Topic	Key Concepts
1	Linear Regression	MSE Loss, Gradient Descent, fit/predict, synthetic data
2	Logistic Regression	Sigmoid, Binary Cross-Entropy, binary classification
3	k-Nearest Neighbors	Euclidean distance, memory, sorting, majority voting
4	Naive Bayes (Gaussian)	Class priors, Gaussian PDF, log prob math
5	Decision Tree	Entropy, Information Gain, recursive tree prediction
6	Random Forest	Bootstrap sampling, ensemble voting
7	Support Vector Machine	Hard-margin linear SVM, max-margin hyperplane
8	Matrix Math Library	Dot products, transpose, scalar ops, operator overloading
9	Perceptron	Basic neural unit, step/tanh, logic gates training
10	Multilayer Perceptron (MLP)	Feedforward, backpropagation, ReLU/sigmoid
11	MNIST with MLP	PGM parsing, feedforward classification
12	Convolutional Neural Networks	2D conv, pooling, stride, combine with MLP
13	Recurrent Neural Networks	Time sequences, memory state, BPTT
14	Optimizers	SGD, Momentum, AdaGrad, RMSProp, Adam
15	Regularization Techniques	L1/L2, Dropout, Early stopping
16	Reinforcement Learning	Q-learning, gridworld, Q-function, Bellman update rule
17	Unsupervised Learning	k-Means clustering, PCA - Principal Component Analysis
18	Genetic Algorithms	Evolutionary algorithms

---

Structure for Each Week

• Theory Markdown (weekX_<topic>.md)
• C++ 20 Implementation:
  • src/models/<topic>.hpp/.cpp
  • tests/test_<topic>.cpp with PASSED/FAILED outputs
• Uses SFML for optional visualization of some of the models like Linear Regression.
• Ephasis on intuition, clarity, and systems-level understanding

---

Final Outcome

• ML library in C++
• Reusable, testable, extendable codebase
• SFML Playground

---

mkdir build
cd build

cmake ...

make