The Inria Rennes - Bretagne Atlantique Centre is one of Inria's eight centres and has more than thirty research teams. The Inria Center is a major and recognized player in the field of digital sciences. It is at the heart of a rich R&D and innovation ecosystem: highly innovative PMEs, large industrial groups, competitiveness clusters, research and higher education players, laboratories of excellence, technological research institute, etc.

The recruittee will work conduct different engineering tasks within the project FAbLe (Framework for Automatic Interpretability in Machine Learning). This project started in 2020 and is financed by a ANR⁽¹⁾ JCJC grant ported by Luis Galárraga, who is a full-time researcher at the LACODAM team. The project is concerned with two major goals: (a) understand what makes an explanation for an AI system understandable, and (b) automate the process of choosing the best explanation depending on the scenario (defined by the AI system, the task, the user's background and expectation, etc).

(1) French National Research Agency

Under the guidance of the FAbLe team , the recruittee's main task will be to implement the algorithms designed and published as an outcome of FAbLe. This implies to provide a Python implementation for the APE framework, that decides whether a black-box classifier on tabular data can be explained accurately and unambigously via a linear approximation.

Supplementary tasks might include:

1. Optimizing the current implementation of the HIPAR algorithm for rule-based regression. This may imply to port some parts of the enumeration phase into Cython
2. Implementation of a simple server interface for a Java in-memory database tailored for rule mining on knowledge graphs.

Main activities (5 maximum) :

• Understand the APE framework, read the scientific papers and look at the current implementation
• Conduct regular meetings with the team members to understand the method and to report on the progress.
• Design, implement, and test the APE framework in Python and scikit-learn
• Release a 0.1 version

Additional activities (3 maximum) :

• Profile the current implementation of the HIPAR algorithm to understand how to optimize
• Optimize the code (potentially using Cython) and release a new version
• Implement a socket-based server interface for the in-memory database used by the AMIE algorithm

Technical skills and level required :

• Ability to code in Python and scikit learn
• Basic knowledge of Machine Learning

Languages :

• English: ability to read and understand scientific articles written in English

• Subsidized meals
• Partial reimbursement of public transport costs
• Possibility of teleworking (90 days per year) and flexible organization of working hours
• Partial payment of insurance costs

monthly gross salary from 2655 euros according to diploma and experience