Program Transformation Lab

We consider source to source transformation as a generic technique in Computer Science. A given / new domain specific language (DSL) is supposed to be translated into a different language X by a DSL2X transformation software. Some additional analyisis / transformation which is not present in the DSL shall be provided by X. The result shall be transformed back into the DSL. We aim for rigorous requirements analysis, object-oriented design and implementation, tests and documentation during the development of the source to source transformation software.

General Information

... on the administrative structure of the lab can be found here.

Topics (subject to last-minute change ...)

  1. ONNX2C[++]
    1. transformation of a feed-forward neural network in ONNX format to C[++] and back
    2. adjoint with dco/c++
    3. visualization with graphviz
  2. ONNX2Python
    1. transformation of a feed-forward neural network in ONNX format to Python and back
    2. adjoint with jax
    3. visualization with graphviz

Agenda

  • Kick-Off: 11.04.2025 at 10am, room 231, IT Center, Seffenter Weg 23.
  • Proposal: Presentation on 25.04.2025 at 11:30am room 231.
  • Requirements and Design: Presentation on 23.05.2025 at 11:30am in room 231.
  • Final Presentation on 04.07.2025 at 11:30am in room 231.
  • Publication (PDF report) by 18.07.2025 at 11:59pm.