• Worked for the Dialogue Systems group at the University of Cambridge, under the supervision of Prof Milica Gašić and Paweł Budzianowski.
  
  
• Examined approaches to extract uncertainty estimates from deep Q-networks (DQN) and implemented deep Bayesian methods for DQN (Bayes-By-Backprop, dropout, concrete dropout, bootstrapped ensemble and alpha-divergences) in PyDial, a Python library for dialogue management.
  
  
• The Bayes-by-Backprop algorithm achieves faster convergence to an optimal policy than any other method, and reaches performance comparable to the state of the art in policy optimization, namely GPSARSA, without the high computational complexity of Gaussian Processes, especially when evaluated on more complex domains.
  
  

Publications

• Tegho, C., Budzianowski, P., & Gašić, M. (2018). Benchmarking Uncertainty Estimates With Deep Reinforcement Learning for Dialogue Policy Optimisation. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP). Paper Link
  
  
• Tegho, C., Budzianowski, P., & Gašić, M. (2017). Uncertainty Estimates for Efficient Neural Network-based Dialogue Policy Optimisation. Accepted at the Bayesian Deep Learning Workshop, 31st Conference on Neural Information Processing Systems (NIPS). Paper Link
  
  
• Tegho, C. Bayes By Backprop Neural Networks forDialogue Management. Thesis Dissertation for MPhil in Machine Learning, Speech and Language Technology, University of Cambridge. Thesis Link
  
  

Scholarships and Awards

• Best Student Paper Award (ICASSP 2018)
• Graduate Masters Scholarship from the Fonds de Recherche - Nature et Technologie Quebec