Machine Learning Kurs im Rahmen der Studierendentage im SS 2023
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
|
|
% Introduction to Data Analysis and Machine Learning in Physics % Martino Borsato, Jörg Marks, Klaus Reygers % 11-14 April 2023 \newline 9:00 - 12:00 and 14:00 - 17:00
## Outline
* **Day 1** - Introduction, software and data fitting
* **Day 2** - Machine learning - basics * **Day 3** - Machine learning - decision tree
* **Day 4** - Machine learning - convolutional networks
* **Organization** and **Objective** - \textcolor{red} {2 ETC: Compulsory attendance is required} \newline \textcolor{red} {Active participation in the exercises} - \textcolor{blue}{Course in CIP pool in a tutorial style} - \textcolor{blue}{Obtain basic knowledge for problem-oriented self-studies} ## Course Information (1)
* Course requirements
- Python knowledge needed / good C++ knowledge might work - Userid to use the CIP Pool of the faculty of physics
* Course structure - \textcolor{red}{Course in CIP pool} using the \textcolor{red}{jupyter3 hub} - Lectures are interleaved with tutorial/exercise sessions in small groups (up to 5 persons / group) * Course homepage which includes and distributes all material \small [https://www.physi.uni-heidelberg.de/~reygers/lectures/2023/ml/](https://www.physi.uni-heidelberg.de/~reygers/lectures/2023/ml/) \normalsize
/transparencies \textcolor{blue}{Transparencies of the lectures} /examples \textcolor{blue}{iPython files shown in the lectures} /exercises \textcolor{blue}{Exercises to be solved during the course} /solutions \textcolor{blue}{Solutions of the exercises}
## Course Information (2)
`TensorFlow` and `Keras` are now also installed in the CIP jupyter hub. In addition, with a google account you can run jupyter notebooks on Google Colab:
\vspace{3ex} [https://colab.research.google.com/](https://colab.research.google.com/)
\vfill
Missing python libraries can be included by adding the following to a cell (here for the pypng library):
``` !pip install pypng ```
## Course Information (3)
* Your installation at home: * \textcolor{blue}{Web Browser to access jupyter3} * \textcolor{blue}{Access to the CIP pool via an ssh client on your home PC} * No requirements for a special operating system
* Software: * firefox or similar * Cisco AnyConnect * ssh client (MobaXterm on Windows, integrated in Linux/Mac)
* Local execution of python / iPython * Install ``anaconda3`` and download / run the iPython notebooks (also python scripts are available)
* \textcolor{red}{Hints for software installations and CIP pool access} \small [https://www.physi.uni-heidelberg.de/~marks/root_einfuehrung/Folien/CIPpoolAccess.PDF](https://www.physi.uni-heidelberg.de/~marks/root_einfuehrung/Folien/CIPpoolAccess.PDF) \normalsize ## Course Information (4)
Alternatively, you can install the libraries needed on your local computer.
\vfill
Here are the relevant instruction for macOS using `pip`:
\vfill
Assumptions: `homebrew` is installed.
\vfill
Install python3 (see https://docs.python-guide.org/starting/install3/osx/) \footnotesize ``` $ brew install python $ python --version Python 3.8.5 ``` \normalsize
Make sure pip3 is up-to-date (alternative: conda $\rightarrow$ don't mix conda and pip installations) \footnotesize ``` $ pip3 install --upgrade pip ``` \normalsize
Install modules needed: \footnotesize ``` $ pip3 install --upgrade jupyter matplotlib numpy pandas scipy scikit-learn xgboost iminuit tensorflow tensorflow_datasets Keras ``` \normalsize
## Topcics and file name conventions
0. Introduction (this file) \hspace{0.1cm} \footnotesize (\textcolor{gray}{introduction.pdf}) \normalsize 1. Introduction to python \hspace{0.1cm} \footnotesize (\textcolor{gray}{01\_intro\_python\_*}) \normalsize 2. Data modeling and fitting \hspace{0.1cm} \footnotesize (\textcolor{gray}{02\_fit\_intro\_*}) \normalsize 3. Machine learning basics \hspace{0.1cm} \footnotesize (\textcolor{gray}{03\_ml\_basics\_*}) \normalsize 4. Decisions trees \hspace{0.1cm} \footnotesize (\textcolor{gray}{04\_decision\_trees\_*}) \normalsize 5. Neural networks \hspace{0.1cm} \footnotesize (\textcolor{gray}{05\_neural\_networks\_*}) \normalsize \vspace{3.5cm}
## Programm Day 1
* Technicalities
* Summary of NumPy
* Plotting with matplotlib
* Input / output of data
* Summary of pandas
* Fitting with iminuit and PyROOT
* Transparencies with activated links, examples and exercises
* Software: [\textcolor{violet}{01\_intro\_python.pdf}](https://www.physi.uni-heidelberg.de/~reygers/lectures/2023/ml/transparencies/01_intro_python.pdf)
* Fitting: [\textcolor{violet}{02\_fit\_intro.pdf}](https://www.physi.uni-heidelberg.de/~reygers/lectures/2023/ml/transparencies/02_fit_intro.pdf) \vspace{2cm}
## Programm Day 2
* Supervised learning
* Classification and regression
* Linear regression
* Logistic regression
* Softmax regression (multi-class classification)
\vspace{4cm}
## Programm Day 3
* Decision trees
* Bagging and boosting
* Random forest
* XGBoost
\vspace{5cm}
## Programm Day 4
* Neural networks
* Convolutional neural networks
* TensorFlow and Keras
* Hand-written digit recognition with Keras
\vspace{5cm}
|