Welcome to the Data Science Practicum 2023 at the Faculty of Science, Masaryk University, during the fall semester of 2023/2024. This course aims to equip you with essential data science skills by delving into various machine learning techniques and datasets through 12 lectures. Rather than focusing solely on theoretical concepts, our emphasis will be on hands-on coding and practical applications.
While the course is still in development, you can refer to its previous versions from [2021] and [2020] to get an idea of the content and structure.
In the upcoming sessions, we will once again cover prominent topics like Convolutional Neural Networks, Transfer Learning, Transformers and NLP methods in Genomics. or the exercises and homework, we will predominantly use Python but it is ok to start with zero knowledge of Python under the condition you can code in another language (e.g. R).
Machine learning has witnessed significant developments in the last two years and we will take advantage of those advancements. Specifically, we will explore diffusion models (used by Stable Diffusion, DALLE and Midjourney) and large language models (behind LlamaChat, Bard, Claude and ChatGPT). Additionally, a key focus of this year's course will be on HuggingFace's transformers and datasets libraries, which have recently become powerhouses driving various neural network applications.
- Lesson 01 (18.9.): Introduction
- Lesson 02 (25.9.): Convolutional neural network (CNN)
- Lesson 03 ( 2.10.): Finetuning
- Lesson 04 ( 9.10.): Hugging Face Spaces
- Lesson 05 (16.10.): Tokenization, Embeddings
- Lesson 06 (23.10.): Transformers, text classification
- Lesson 07 (30.10.): Interpretability
- Lesson 08 ( 6.11.): Exercises: table of astronauts
- Lesson 09 (13.11.): Large language models
- Lesson 10 (20.11.): Exercises: ML pipeline and Titanic dataset
- Lesson 11 (27.11.): Stable Diffusion