Repository "ML/AI in Julia Course" at TU Kaiserslautern.
The challenge of bringing projects from research to real-world impact is spinning out of control. Ideas need to reach the cloud and clusters for large-scale data for real-time analysis and be built on by other industrial/ academic projects. Ad-hoc solutions by fragmented research groups using different languages create sprawling codebases and myriad tools that impair collaboration and innovation. Meanwhile, the well-oiled machines of Google (Deepmind), Facebook and others loom large over academic initiatives.
The Julia language offers a compelling solution: A single system (no need for specialist wrapping C++ code - Julia code runs on GPUs! - single code base for end-user) that can meet all of these demands at once, cutting down the innovation cycle by months or years. Julia can easily tie in to pre-existing solution in any language. The general user can implement non-standard variations of common algorithms without the need to write and understand low-level code - Moreover, Additional arguments for Julia are performance and parallel computing support in contrast to dominant languages R and Python.
This course will introduce the language and demonstrates Machine Learning and Probabilistic Programming in that context.
Please install Julia as explained at https://computationalthinking.mit.edu/Spring21/installation/ but Version 1.7.
- Introduction to Julia
- Handling and Exploring Data
- Statistics and Machine Learning
- Neural Networks in Flux.jl
- Bayesian Inference and a Glimpse of Hybrid Model
- Probabilistic Programming in Omega.jl and Turing.jl
-
Julia for Python people https://colab.research.google.com/drive/1G04w8tTl074180DP_Ka9X44r_pndUYxq?usp=sharing
-
Julia for Machine Learning, Zacharias Voulgaris, https://technicspub.com/julia/
-
N. D. Goodman, J. B. Tenenbaum, and The ProbMods Contributors (2016). Probabilistic Models of Cognition (2nd ed.). Retrieved 2022-1-6from https://probmods.org/ And its upcoming translation into Julia https://github.com/zenna/Omega.jl
Requirements for attendance (formal) ML1