SCIENTIFIC COMMITTEE

Syllabus

Topics that tasks may cover in every round of the 2027 edition. Tasks do not require knowledge beyond this list.

The Syllabus is prepared by the Scientific Committee and may be updated between seasons. Changes are published on iaitc-official.org before the IAITC Open I.

01

Reinforcement Learning and AI Search

  1. A* algorithm and heuristics
  2. Minimax and variations
  3. Monte Carlo method
  4. Markov Decision Processes (MDPs)
  5. Temporal Difference learning
  6. Q-learning
02

Machine Learning

  1. Classical algorithms (classification and regression): Naive Bayes, Linear Regression, Logistic Regression, SVM, k-NN, Decision Trees
  2. Clustering: K-Means, K-Means++, DBSCAN, Hierarchical, Spectral
  3. Ensemble methods: Random Forests, Voting, Bagging, Boosting
  4. Dimensionality reduction
03

Deep Learning

  1. Neural network basics: Perceptron, MLP, Backpropagation
  2. Optimisation techniques: SGD, Adam, RMSProp, learning rate schedules, regularisation, dropout, batch normalisation
  3. Architectures: CNNs, RNNs, LSTM, GRU, Transformers (attention mechanisms, BERT, GPT, Vision Transformers)
  4. Generative models: Autoencoders, Variational Autoencoders (VAE), GANs, Diffusion models
04

Natural Language Processing

  1. Preprocessing: tokenisation, stemming, etc.
  2. Embeddings: TF-IDF, Word2Vec, etc.
  3. Related architectures: Seq2Seq, T5, LLaMA, etc.
05

Computer Vision

  1. Processing: filtering, edge detection, Histogram of Oriented Gradients (HOG), etc.
  2. CNN architectures: AlexNet, VGG, ResNet, Inception, EfficientNet
  3. Related architectures: YOLO, Stable Diffusion, Vision Transformers (ViT), etc.
06

Audio Processing

  1. Representation: audio as NumPy arrays, spectrogram, mel-spectrogram
  2. Augmentations: time shift, pitch shift, speed change
  3. Pre-trained audio encoders: HuBERT
  4. Audio models: Qwen-Audio, Whisper, Voxtral
Some tasks are “From Scratch” tasks, which test theoretical and algorithmic depth. They must be solved with NumPy, pandas and the Python Standard Library only, without libraries such as scikit-learn, PyTorch, TensorFlow or OpenCV.