AI learning path

Build a neural network from scratch in C++

15 challenges take you from the dot product to a network that learns XOR, then on to modern optimizers like Adam, with no AI libraries: neurons, activations, loss, gradients, backpropagation and training. Solve them in order in C++; each one runs in your browser with instant tests and an AI review.

Start the first challenge

0%solved

0 / 15

1

Building blocks

Vectors, activation functions and a single neuron.

0/3
Learn the concept first Neurons and Activation Functions in Python Read the lesson →
1. 1 Dot Product of Two Vectors Easy Solve in C++
2. 2 Activation Functions Easy Solve in C++
3. 3 Single Artificial Neuron Easy Solve in C++
2

Forward pass

Layers, a two-layer network and loss functions.

0/3
Learn the concept first Forward Propagation and Loss Functions in Python Read the lesson →
1. 1 Dense Layer Forward Pass Medium Solve in C++
2. 2 Forward Propagation in a Two-Layer Network Medium Solve in C++
3. 3 Loss Functions: MSE, MAE and Cross-Entropy Easy Solve in C++
3

Training

Gradients, backpropagation and gradient descent, up to solving XOR.

0/5
Learn the concept first Gradient Descent and Backpropagation in Python Read the lesson →
1. 1 Derivatives of Activation Functions Easy Solve in C++
2. 2 Gradients of a Single Neuron Medium Solve in C++
3. 3 Backpropagation in a Two-Layer Network Hard Solve in C++
4. 4 Train a Neuron with Gradient Descent Medium Solve in C++
5. 5 Train a Neural Network to Solve XOR Hard Solve in C++
4

Better training

Weight initialization, mini-batches, momentum and Adam.

0/4
Learn the concept first Neural Network Optimizers: SGD, Momentum and Adam in Python Read the lesson →
1. 1 Xavier and He Weight Initialization Easy Solve in C++
2. 2 Mini-Batch Gradient Descent Medium Solve in C++
3. 3 Gradient Descent with Momentum Easy Solve in C++
4. 4 Adam Optimizer Medium Solve in C++