From b7dd24b5fa2a86d341e5c73843ef41a8e79204c1 Mon Sep 17 00:00:00 2001 From: Matt McKay Date: Sat, 1 Aug 2026 09:29:06 +1000 Subject: [PATCH 1/2] Update translation: lectures/mccall_q.md --- lectures/mccall_q.md | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/lectures/mccall_q.md b/lectures/mccall_q.md index 02a1a281..801db5e0 100644 --- a/lectures/mccall_q.md +++ b/lectures/mccall_q.md @@ -90,7 +90,7 @@ FONTPATH = "fonts/SourceHanSerifSC-SemiBold.otf" mpl.font_manager.fontManager.addfont(FONTPATH) plt.rcParams['font.family'] = ['Source Han Serif SC'] -np.random.seed(123) +rng = np.random.default_rng(123) ``` ## McCall 模型回顾 @@ -499,15 +499,15 @@ class Qlearning_McCall: self.quit_allowed = quit_allowed - def draw_offer_index(self): + def draw_offer_index(self, rng): """ 从工资分布中抽取状态索引。 """ q = self.q - return np.searchsorted(np.cumsum(q), np.random.random(), side="right") + return np.searchsorted(np.cumsum(q), rng.random(), side="right") - def temp_diff(self, qtable, state, accept): + def temp_diff(self, qtable, state, accept, rng): """ 计算与状态和动作相关的TD。 """ @@ -515,7 +515,7 @@ class Qlearning_McCall: c, β, w = self.c, self.β, self.w if accept==0: - state_next = self.draw_offer_index() + state_next = self.draw_offer_index(rng) TD = c + β*np.max(qtable[state_next, :]) - qtable[state, accept] else: state_next = state @@ -526,7 +526,7 @@ class Qlearning_McCall: return TD, state_next - def run_one_epoch(self, qtable, max_times=20000): + def run_one_epoch(self, qtable, rng, max_times=20000): """ 运行一个"轮次"。 """ @@ -534,7 +534,7 @@ class Qlearning_McCall: c, β, w = self.c, self.β, self.w eps, δ, lr, T = self.eps, self.δ, self.lr, self.T - s0 = self.draw_offer_index() + s0 = self.draw_offer_index(rng) s = s0 accept_count = 0 @@ -542,7 +542,7 @@ class Qlearning_McCall: # 选择动作 accept = np.argmax(qtable[s, :]) - if np.random.random()<=eps: + if rng.random()<=eps: accept = 1 - accept if accept == 1: @@ -550,7 +550,7 @@ class Qlearning_McCall: else: accept_count = 0 - TD, s_next = self.temp_diff(qtable, s, accept) + TD, s_next = self.temp_diff(qtable, s, accept, rng) # 更新qtable qtable_new = qtable.copy() @@ -567,7 +567,7 @@ class Qlearning_McCall: return qtable_new @jit -def run_epochs(N, qlmc, qtable): +def run_epochs(N, qlmc, qtable, rng): """ 运行N次轮次，每次使用上一次迭代的qtable。 """ @@ -575,7 +575,7 @@ def run_epochs(N, qlmc, qtable): for n in range(N): if n%(N/10)==0: print(f"进度：轮次 = {n}") - new_qtable = qlmc.run_one_epoch(qtable) + new_qtable = qlmc.run_one_epoch(qtable, rng) qtable = new_qtable return qtable @@ -593,7 +593,7 @@ qlmc = Qlearning_McCall() # 运行 qtable0 = np.zeros((len(w_default), 2)) -qtable = run_epochs(20000, qlmc, qtable0) +qtable = run_epochs(20000, qlmc, qtable0, rng) ``` ```{code-cell} ipython3 @@ -670,7 +670,7 @@ def plot_epochs(epochs_to_plot, quit_allowed=1): ax.plot(w_new, valfunc_qlr, '-o', label=f'QL:训练轮数={n}, 平均误差={error}') - new_qtable = qlmc_new.run_one_epoch(qtable) + new_qtable = qlmc_new.run_one_epoch(qtable, rng) qtable = new_qtable ax.set_xlabel('工资') From 6d9612b6955aec7b6cc845be43b4e18efceb5c22 Mon Sep 17 00:00:00 2001 From: Matt McKay Date: Sat, 1 Aug 2026 09:29:07 +1000 Subject: [PATCH 2/2] Update translation: .translate/state/mccall_q.md.yml --- .translate/state/mccall_q.md.yml | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/.translate/state/mccall_q.md.yml b/.translate/state/mccall_q.md.yml index ba6a9118..35b9a5bb 100644 --- a/.translate/state/mccall_q.md.yml +++ b/.translate/state/mccall_q.md.yml @@ -1,6 +1,6 @@ -source-sha: 74949c4928e482c39af61ab3577300621d262f37 -synced-at: "2026-07-18" +source-sha: 4a5fad71a60d8467265104ec0fdaf51cd5252966 +synced-at: "2026-07-31" model: claude-sonnet-5 -mode: RESYNC +mode: UPDATE section-count: 7 -tool-version: 0.17.0 +tool-version: 0.24.0