1 Experiments

Nothing again. I need to start avoiding a lot of managerial drudgery in work. I think there are a few steps I should start taking:

  1. Enforce hours and smaller time spans for meetings not related to actual ML work.
  2. Start using simpler tasks management systems wherever needed. I don't mind sticky notes at this moment to be honest. JIRA can fuck right off. Real world project management is so much bike shedding.
  3. Another process eating my time is hiring. I probably should start automating loads of that by structuring the evaluation methodologies.

Seeking perfection in organizational tasks is doomed to screw my real directions of work.

2 Readings

3 Programming

Commits for week 50-2019 and 4 previous weeks.


  • [bisani2004bootstrap] "Bisani & Ney". 2004. "Bootstrap estimates for confidence intervals in ASR performance evaluation", 409-412, in in: "2004 IEEE International Conference on Acoustics, Speech, and Signal Processing", edited by
  • [conway1968committees] Conway. 1968. "How do committees invent." Datamation, 14(4), 28-31. link. doi.
  • [dalmia2019phoneme] "Siddharth Dalmia, Xinjian Li, Alan W Black & Florian Metze". 2019. "Phoneme Level Language Models for Sequence Based Low Resource ASR", 6091-6095, in in: "ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)", edited by
  • [grathwohl2019classifier] @miscgrathwohl2019classifier, Author = Will Grathwohl and Kuan-Chieh Wang and Jörn-Henrik Jacobsen and David Duvenaud and Mohammad Norouzi and Kevin Swersky, Title = Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One, Year = 2019, Eprint = arXiv:1912.03263,