Melih Kandemir


I joined Bosch Center for Artificial Intelligence. This website will no longer be maintained.

I used to be an assistant professor at Özyeğin UniversityComputer Science Department. Back then, my research activity focused primarily on Bayesian inference of deep neural nets, transfer learning, meta-learning, few-shot learning, reinforcement learning, and application of these approaches to computer vision and medical image analysis problems. Currently I still do similar things.

E-mail: name [dot] surname [at] ozyegin [dot] edu [dot] tr
Address: Nişantepe mah. Orman sok. 34-36, Alemdağ, Çekmeköy, İstanbul
Office: EF 104 | Phone: +90-216-564-9537
Scholar | GitHub | LinkedIn


  • 02.01.2018– Gave an interview to Para, a popular economics magazine in Turkey, about applications of ML to histopathology cancer diagnostics. The link is here.
  • 24.11.2017– Our paper at NoF 2017 has received the Best Paper Award.
  • 19.10.2017- Received TÜBİTAK 2232 Career Reintegration Grant
  • 21.08.2017- Gave an invited talk at the Turkish Machine Learning Summer School 2017 on probabilistic deep learning.
  • 09.08.2017- Gave an interview to the pioneer AI company of the Turkish market, Etiya. The link is here.



  • Sampling-free Variational Inference of Bayesian Neural Nets
    M. Kandemir, M. Haußmann, F.A. Hamprecht [PDF]


(Full list available here)

  • Evidential Deep Learning to Quantify Classification Uncertainty
    M. Sensoy, L. Kaplan, M. Kandemir
    NIPS, (2018) [PDF]
  • Variational Closed-Form Deep Neural Net Inference
    M. Kandemir
    Pattern Recognition Letters, (2018), [PDF]
  • On context aware DDoS attacks using deep generative networks
    G. Gürsun, M. Şensoy, M. Kandemir
    ICCCN, (2018) 
  • Supervising topic models with Gaussian processes
    M. Kandemir, T. Kekeç, R. Yeniterzi
    Pattern Recognition, (2018) [PDF]
  • Prediction of active UE number with Bayesian neural networks for self-organizing LTE networks
    Ö. Narmanlıoğlu, E. Zeydan, M. Kandemir, T. Kranda,
    Network of the Future (NoF), Proceedings, (2017) [PDF]
    Best Paper Award
  • Variational Bayesian multiple instance learning with Gaussian processes
    M. Haußmann, F.A. Hamprecht, M. Kandemir
    CVPR, Proceedings, (2017)  [PDF]


  • Gaussian process density counting from weak supervision
    M. von Borstel, M. Kandemir, P. Schmidt, M. Rao, K. Rajamani, F.A. Hamprecht
    ECCV, Proceedings, (2016) [PDF]
  • Asymmetric transfer learning with deep Gaussian processes
    M. Kandemir
    ICML, Proceedings, (2015) [PDF] [Code][Talk]
  • Towards brain-activity-controlled information retrieval: Decoding image relevance from MEG signals
    J.P. Kauppi *, M. Kandemir *, V.M. Saarinen, L. Hirvenkari, L. Parkkonen, A.Klami, R. Hari, S. Kaski
    NeuroImage, (2015) * Equal Contributions [PDF]
  • Instance label prediction by Dirichlet process multiple instance learning
    M. Kandemir, F.A. Hamprecht
    UAI, Proceedings, (2014) [PDF] [Code]