def update_trial_learning_cos_diff(self) -> None:
cos_diff = torch.nn.functional.cosine_similarity(
self.post.phase_acts[self.plus_phase],
self.post.phase_acts[self.minus_phase],
dim=0)
self.cos_diff = utils.clip_float(low=0.01, high=0.99, x=cos_diff)
self.cos_diff_avg = self.post.spec.avg_dt * (
cos_diff - self.cos_diff_avg)
def update_trial_learning_averages(self) -> None:
"""Updates the learning averages computed at the end of each trial."""
cos_diff = torch.nn.functional.cosine_similarity(
self.acts_p, self.acts_m, dim=0)
self.cos_diff = utils.clip_float(low=0.01, high=0.99, x=cos_diff)
self.cos_diff_avg += self.spec.avg_dt * (cos_diff - self.cos_diff_avg)
acts_p_avg_eff = self.acts_p.mean().item()
self.units.update_trial_learning_averages(acts_p_avg_eff)
def test_clip_float_returns_the_lower_bound_if_x_is_below_it() -> None:
assert ut.clip_float(low=0.2, high=0.5, x=-1) == 0.2
def test_clip_float_returns_x_if_it_is_within_the_range() -> None:
assert ut.clip_float(low=0.2, high=0.5, x=0.3) == 0.3
def test_clip_float_returns_the_upper_bound_if_x_is_above_it() -> None:
assert ut.clip_float(low=0.2, high=0.5, x=1) == 0.5