Poor Machine Learning System
- 92% on the ML task is probably not very good.
- Eliding the lower formants moved it to 98%.
- I'm pretty sure I could have come up with some statistical method,
that would have got well above 99%, and I haven't looked
at how others have done this task (if they do it at all).
- I'd like my neural systems to actually be leading edge ML systems,
and this isn't.
- However, the point of this work is not (at least yet) to have
leading edge speech recognition.
- We'd like to build a good neuro-cognitive model of the task.
