Postdoc position on generative modelling for complex objects (closed)

We are looking for an excellent postdoc to work on the development of deep learning methods for the automatic composition of complex structures, such as sets, graphs, trees, sequences, that exhibit desired properties. Typical application scenarios include image and text generation, drug and molecule design. Research areas of direct interest include generative modeling, unsupervised and semi-supervised learning.

Recent advances in generative modelling have made possible the modelling of complex distributions, such as images, which then allows one to draw samples from the model and generate quite realistic samples, 1, 2, 3. We want to build on these and similar works to learn conditional generative models which will allow us to sample structures with desired properties. Such ideas and approaches have recently emerge in the field of molecule design and generation 4, 5, as well as in text generation 6. In addition we want the developed methods to be able to exploit large collections of non annotated structures, i.e. they should be applicable in semi-supervised settings.

The position is funded for two years. Annual gross salary, depending on qualifications, up to 92kCHF.

The successful candidate will join the data mining and machine learning team of the University of Applied Sciences, Western Switzerland, led by Prof. Alexandros Kalousis, and will also be associated with the Computer Science department of the University of Geneva within the VIPER group led by Prof. Stephane Marchand-Maillet. Our research explores a number of different issues such as: learning in high dimensional settings, dimensionality reduction and feature selection, learning with structured data (multiple kernel learning), metric and similarity learning, the exploitation of domain knowledge in the learning process, and learning over dynamic/temporal data. For a more detailed description the interested candidates may take a look at: and the list of publications within there. The greater Geneva lake area is a world-renowned education and research hub, including not only the University of Geneva, but also EPFL, and IDIAP. It offers considerable opportunities for training and exposure to machine learning, with a number of research teams being active on these and related fields. In addition the selected candidate will have ample opportunities to participate in the main ML and DM conferences.

The ideal candidate will have:

  • A PhD on machine learning, or other strongly related discipline.
  • A very solid background in a combination of computer science and mathematics. Special areas of interest include: deep learning, statistical machine learning, mathematical optimization .
  • Strong publication record in the area of machine learning and data mining (e.g. ICML, NIPS, KDD, IDCM etc).
  • Experience in generative models will be a considerable plus.
  • Solid expertise in at least one of the standard deep learning packages such as TensorFlow, Theano, Torch.
  • Excellent command of English.
  • Team work capacity.

Candidates should send:

  • A two page CV.
  • A one page motivation letter explaining why their skills, knowledge and experience make them a particularly suitable candidate for the given position.
  • A 1000 words research proposal on deep learning methods for the generation of complex structures with given properties.
  • Their three most representative papers.
  • The contact details of three referees; do not send reference letters.

to Application Deadline Priority will be given to applications received by the 15/Sept/2017, however applications will be accepted until the position is filled. The position status will be indicated here.

The position is available ASAP.


  1. Generative Adversarial Networks, Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, Yoshua Bengio,
  2. Adversarial Autoencoders, Alireza Makhzani, Jonathon Shlens, Navdeep Jaitly, Ian Goodfellow, Brendan Frey,
  3. Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space, Anh Nguyen, Jeff Clune, Yoshua Bengio, Alexey Dosovitskiy, Jason Yosinski,
  4. Automatic chemical design using a data-driven continuous representation of molecules, Rafael Gomez-Bombarelli, David Duvenaud, Jose Miguel Hern√°ndez-Lobato, Jorge Aguilera-Iparraguirre, Timothy D. Hirzel, Ryan P. Adams, Alan Aspuru-Guzik,
  5. . Towards "AlphaChem": Chemical Synthesis Planning with Tree Search and Deep Neural Network Policies, Marwin Segler, Mike Preuß, Mark P. Waller,
  6. Controllable Text Generation, Zhiting Hu, Zichao Yang, Xiaodan Liang, Ruslan Salakhutdinov, Eric P. Xing,