[CozySocksAlways]

Description

• Clarifies how the dither parameter of trial_shifting works and makes the code match the documented behaviour.
• Old docstring for dither fails to mention:
  • If the shift is applied to the whole trial, i.e that each spike is moved by same amount
  • If the shift can go in either direction up to dither
  • What distribution the shift is drawn from (uniform)

Changes

• Docs (trial_shifting, _trial_shifting_of_concatenated_spiketrain): the dither entry now says that all spikes in a trial are shifted by the same random amount, drawn uniformly from the closed interval [-dither, +dither].
• Implementation (_trial_shifting): the shift is now drawn with random.uniform(-dither, dither) instead of dither * (2 * random.random() - 1). This also matches the intent mentioned in the documentation and also reads better in code. The behavior is unchanged, so the distribution is same.

[coveralls]

coverage: 88.242% (-0.4%) from 88.605% — INM-6:fix/trial_shift_doc into NeuralEnsemble:master