digital-garbageandClaude Sonnet 5.5
f41136e3e0
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
60 lines
2.4 KiB
Python
60 lines
2.4 KiB
Python
"""One number per run, learned from Liked/Disliked ratings.
|
|
|
|
Each run tries a value near the key's current centre; the rating of that run pulls the
|
|
centre toward the tried value (liked) or pushes it away (disliked). A rating counts more
|
|
when the run's measured effect was unusually strong for this key and less when unusually
|
|
weak: a verdict on a run that barely felt the feature says little about it.
|
|
|
|
Pure arithmetic over rated rows. The caller keeps the rows (the taste store) and fills a
|
|
capture `{"v": tensor, "e": tensor}` during the run; this file holds no state.
|
|
"""
|
|
|
|
import random
|
|
|
|
MIN_EFFECTS = 3 # runs of history before an effect is compared with the key's own
|
|
|
|
|
|
def history(rows, setup=None):
|
|
"""Taste-store rows -> [(value, reward, effect | None)], oldest first.
|
|
|
|
`setup` is what gives the value its meaning ({"b": block}): a rating made with another
|
|
setup says nothing about this one and is left out. Rows from before a setup was
|
|
recorded are left out too."""
|
|
out = []
|
|
for r in rows:
|
|
p = r["rows"]
|
|
if "v" in p and all(k in p and int(p[k]) == int(want) for k, want in (setup or {}).items()):
|
|
e = float(p["e"]) if "e" in p and float(p["e"]) > 0 else None
|
|
out.append((float(p["v"]), float(r["reward"]), e))
|
|
return out
|
|
|
|
|
|
def unusualness(effect, past, lo=0.25, hi=2.0):
|
|
"""Weight of one rating: this run's effect over the key's median past effect, clamped.
|
|
1.0 until enough runs exist to know what usual is."""
|
|
if not effect or len(past) < MIN_EFFECTS:
|
|
return 1.0
|
|
med = sorted(past)[len(past) // 2]
|
|
return min(max(effect / med, lo), hi) if med > 0 else 1.0
|
|
|
|
|
|
class Dial:
|
|
def __init__(self, start, lo, hi, explore, lr=0.5):
|
|
self.start, self.lo, self.hi, self.explore, self.lr = start, lo, hi, explore, lr
|
|
|
|
def centre(self, hist):
|
|
m, seen = self.start, []
|
|
for v, reward, effect in hist:
|
|
sign = (reward > 0) - (reward < 0)
|
|
m += self.lr * sign * unusualness(effect, seen) * (v - m)
|
|
m = min(max(m, self.lo), self.hi)
|
|
if effect:
|
|
seen.append(effect)
|
|
return m
|
|
|
|
def pick(self, hist, rng=random):
|
|
"""-> (value to try, centre, ratings held)."""
|
|
centre = self.centre(hist)
|
|
v = min(max(centre + rng.gauss(0.0, self.explore), self.lo), self.hi)
|
|
return v, centre, len(hist)
|