菜单
Shape RL rewards with time-decaying asset weights and time-increasing resource weights so the agent transitions from expansion to accumulation as the game progresses
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Memory-efficient Keras generator that streams sharded CSV files in chunks, renders strokes to images on-the-fly, and yields batches for training on datasets too large for memory
Stack 1D convolutions for local feature extraction before bidirectional LSTMs to classify variable-length stroke sequences into hundreds of doodle categories
Partition a large dataset into N balanced shards using integer key modulo arithmetic for reproducible, class-interleaved splits across CSV files
Normalize raw stroke coordinates to 0-255 range, resample at uniform arc-length spacing, then apply Ramer-Douglas-Peucker simplification
Render stroke sequences to grayscale images with temporal intensity encoding where earlier strokes are brighter and later strokes fade to encode drawing order
Score each face in a video as anomalous by computing L2 distance from the embedding centroid of all faces, then convert to probability via logistic function
| name | tabular-time-varying-reward-shaping |
| description | Shape RL rewards with time-decaying asset weights and time-increasing resource weights so the agent transitions from expansion to accumulation as the game progresses |
In resource-management games, the optimal strategy shifts over time: early game rewards fleet/structure building (assets), late game rewards resource hoarding. Time-varying reward shaping encodes this by linearly interpolating weights between assets and resources as a function of turn number, plus a terminal win bonus inversely proportional to game length.
import numpy as np
max_steps = 400
transition_point = 300
w_assets = np.concatenate([
np.linspace(1.0, 0.0, transition_point),
np.zeros(max_steps - transition_point)
])
w_resources = np.concatenate([
np.linspace(0.0, 1.0, transition_point),
np.ones(max_steps - transition_point)
])
def board_value(player, turn, ship_cost=10, yard_cost=50):
val_resources = w_resources[turn] * player.kore
val_ships = w_assets[turn] * ship_cost * player.total_ships
val_yards = w_assets[turn] * yard_cost * player.total_shipyards
return val_resources + val_ships + val_yards
def reward(board, prev_board, turn, done, won):
r = board_value(board.current_player, turn) - board_value(prev_board.current_player, turn - 1)
if done:
bonus = (1 if won else -1) * (100 + 5 * (max_steps - turn))
r += bonus
return r
remaining_turns * constant encourages decisive play over stalling