🐍 AI新手必學的 Python技巧

# 整數 / 浮點數 / 字串 / 布林
age     = 25
score   = 98.5
name    = "小明"
is_ai   = True

# 查看型別
print(type(age))    # <class 'int'>
print(type(score))  # <class 'float'>
print(type(name))   # <class 'str'>

# 型別轉換
num_str = str(42)        # "42"
num_int = int("100")     # 100
num_f   = float("3.14") # 3.14

name  = "Claude"
score = 95

# f-string（推薦寫法）
msg = f"模型 {name} 得分：{score} 分"

# 常用字串方法
text = "  Hello, AI World!  "
print(text.strip())        # 去空白
print(text.lower())        # 全小寫
print(text.upper())        # 全大寫
print(text.replace("AI", "人工智慧"))

# 切割與合併
words = "Python,NumPy,Pandas".split(",")
joined = " | ".join(words)
print(words)   # ['Python', 'NumPy', 'Pandas']
print(joined)  # Python | NumPy | Pandas

# 字串切片
s = "AI Python"
print(s[0:2])   # "AI"
print(s[-6:])  # "Python"

confidence = 0.87

if confidence >= 0.9:
    print("高信心度預測")
elif confidence >= 0.7:
    print("中等信心度，建議人工確認")
else:
    print("低信心度，請重新輸入")

# 單行三元運算式
label = "貓" if confidence > 0.5 else "不確定"

# 多條件組合
is_valid = confidence > 0.6 and confidence <= 1.0
print(f"結果有效：{is_valid}")

models = ["GPT-4", "Claude", "Gemini"]

# 基本 for 迴圈
for model in models:
    print(f"測試模型：{model}")

# enumerate — 同時取索引與值（AI常用）
for i, model in enumerate(models):
    print(f"[{i}] {model}")

# range — 重複執行
for epoch in range(1, 6):
    print(f"訓練第 {epoch} 輪")

# zip — 同時遍歷兩個清單
scores = [92, 88, 95]
for model, score in zip(models, scores):
    print(f"{model}: {score}分")

# while 迴圈
attempt = 0
while attempt < 3:
    print(f"第 {attempt+1} 次嘗試")
    attempt += 1

data = [10, 20, 30, 40, 50]

# 新增 / 刪除
data.append(60)       # 尾端新增
data.extend([70, 80]) # 合併另一個 list
data.pop()            # 移除最後一個
data.pop(0)          # 移除第一個 (index 0)

# 切片
print(data[0:3])   # 前3筆
print(data[-3:])  # 後3筆
print(data[::2])  # 每隔一個取

# 排序
data.sort()               # 升冪排序
data.sort(reverse=True)  # 降冪排序

# List Comprehension（AI常用）
squares = [x**2 for x in range(10)]
evens   = [x for x in range(20) if x % 2 == 0]
print(squares)  # [0, 1, 4, 9, 16, ...]
print(evens)    # [0, 2, 4, 6, ...]

# 建立字典
model_config = {
    "model": "claude-sonnet-4-6",
    "temperature": 0.7,
    "max_tokens": 1024,
    "top_p": 0.9
}

# 存取值
print(model_config["model"])           # 直接存取
print(model_config.get("temperature")) # 安全存取（不存在不報錯）
print(model_config.get("seed", 42))   # 帶預設值

# 新增 / 修改
model_config["stream"] = True
model_config.update({"temperature": 0.5, "seed": 42})

# 遍歷
for key, value in model_config.items():
    print(f"  {key}: {value}")

# Dict Comprehension
scores = {"GPT-4": 92, "Claude": 95, "Gemini": 88}
high = {k: v for k, v in scores.items() if v > 90}
print(high)  # {'GPT-4': 92, 'Claude': 95}

# Tuple — 不可變的串列
image_size = (224, 224, 3)  # 影像 (H, W, C)
h, w, c = image_size          # 解包（unpacking）
print(f"高:{h} 寬:{w} 通道:{c}")

# 函數回傳多值也用 Tuple
def get_stats(data):
    return min(data), max(data), sum(data)/len(data)

mn, mx, avg = get_stats([10, 20, 30])

# Set — 去重與集合運算
labels_a = {"貓", "狗", "鳥"}
labels_b = {"狗", "魚", "鳥"}

print(labels_a & labels_b)  # 交集: {'狗', '鳥'}
print(labels_a | labels_b)  # 聯集: {'貓','狗','鳥','魚'}
print(labels_a - labels_b)  # 差集: {'貓'}

# 去重
tags = ["AI", "ML", "AI", "DL", "ML"]
unique = list(set(tags))  # ['AI', 'ML', 'DL']

# 基本函數
def greet(name: str, lang: str = "zh") -> str:
    if lang == "zh":
        return f"你好，{name}！"
    return f"Hello, {name}!"

print(greet("小明"))           # 你好，小明！
print(greet("Alice", "en")) # Hello, Alice!

# *args — 不定數量位置參數
def average(*nums):
    return sum(nums) / len(nums)

print(average(80, 90, 100))  # 90.0

# **kwargs — 不定數量關鍵字參數
def build_prompt(**kwargs):
    return {"messages": [kwargs]}

# lambda — 單行匿名函數
normalize = lambda x, mx: x / mx
print(normalize(75, 100))  # 0.75

# 標準庫
import os
import json
import datetime
from pathlib import Path

# 常用操作
cwd = os.getcwd()           # 目前目錄
now = datetime.datetime.now()
files = list(Path(".").glob("*.py"))

# JSON — API 回應解析常用
data = {"model": "claude", "tokens": 1024}
json_str = json.dumps(data, ensure_ascii=False)
parsed   = json.loads(json_str)
print(parsed["model"])  # claude

# AI 常用套件安裝指令：
# pip install numpy pandas matplotlib scikit-learn
# pip install anthropic openai torch transformers

# 別名導入（常見慣例）
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

scores = [55, 72, 88, 43, 95, 61]

# map — 對每個元素套用函數
normalized = list(map(lambda x: x/100, scores))
# [0.55, 0.72, 0.88, 0.43, 0.95, 0.61]

# filter — 過濾符合條件的元素
passing = list(filter(lambda x: x >= 60, scores))
# [72, 88, 95, 61]

# sorted — 自訂排序鍵
models = [
    {"name": "GPT-4", "score": 92},
    {"name": "Claude", "score": 95},
    {"name": "Gemini", "score": 88},
]
ranked = sorted(models, key=lambda m: m["score"], reverse=True)
for r in ranked:
    print(f"{r['name']}: {r['score']}")
# Claude: 95 / GPT-4: 92 / Gemini: 88

import numpy as np

# 建立 Array
a = np.array([1, 2, 3, 4, 5])
b = np.zeros((3, 4))          # 3x4 全零矩陣
c = np.ones((2, 3))           # 2x3 全一矩陣
d = np.arange(0, 10, 2)       # [0, 2, 4, 6, 8]
e = np.linspace(0, 1, 5)      # 0到1均勻分5點
r = np.random.rand(3, 3)     # 3x3 隨機矩陣

# 形狀操作（AI最重要！）
x = np.arange(12)
x2d = x.reshape(3, 4)  # 變成 3x4
x4d = x.reshape(2, 2, 3)

print(x2d.shape)   # (3, 4)
print(x2d.dtype)   # int64
print(x2d.ndim)    # 2（二維）
print(x2d.size)    # 12（元素總數）

import numpy as np

a = np.array([1, 2, 3, 4])
b = np.array([10, 20, 30, 40])

# 向量化 — 不需要迴圈
print(a + b)   # [11, 22, 33, 44]
print(a * 2)   # [2, 4, 6, 8]（廣播）
print(a ** 2)  # [1, 4, 9, 16]

# 統計函數（AI訓練監控必用）
data = np.array([85, 90, 78, 92, 88])
print(np.mean(data))    # 平均值 86.6
print(np.std(data))     # 標準差
print(np.max(data))     # 最大值 92
print(np.argmax(data))  # 最大值的索引 3（AI分類用）

# 矩陣乘法（神經網路的核心運算）
W = np.random.rand(3, 4)  # 權重矩陣
x = np.random.rand(4)     # 輸入向量
y = W @ x                  # 矩陣乘法，等同 np.dot(W, x)
print(y.shape)  # (3,)

# Softmax（分類輸出）
logits = np.array([2.1, 0.5, 1.3])
exp = np.exp(logits)
probs = exp / np.sum(exp)
print(probs)  # 和為 1 的機率分佈

import pandas as pd

# 讀取 CSV（最常見操作）
df = pd.read_csv("data.csv")

# 建立 DataFrame
df = pd.DataFrame({
    "name":   ["Alice", "Bob", "Carol"],
    "score":  [92, 85, 95],
    "passed": [True, True, True]
})

# 基本查看
print(df.head(3))    # 前3行
print(df.tail(2))    # 後2行
print(df.shape)      # (3, 3) → 3列3欄
print(df.dtypes)     # 各欄型別
print(df.describe()) # 統計摘要
print(df.info())     # 完整資訊含缺失值

# 存取欄位
scores = df["score"]        # Series
subset = df[["name", "score"]]  # 多欄 DataFrame

import pandas as pd
import numpy as np

df = pd.DataFrame({
    "age":   [25, np.nan, 30, 22, np.nan],
    "city":  ["台北", "台中", None, "高雄", "台北"],
    "score": [85, 90, 78, 92, 88]
})

# 處理缺失值
print(df.isnull().sum())       # 各欄缺失數量
df_clean = df.dropna()         # 刪除含缺失值的行
df["age"].fillna(df["age"].mean(), inplace=True)  # 用平均值填補

# 條件篩選
high = df[df["score"] > 85]
taipei = df[df["city"] == "台北"]
multi  = df[(df["score"] > 80) & (df["age"] < 30)]

# 群組統計
grouped = df.groupby("city")["score"].mean()
print(grouped)

# 新增欄位
df["grade"] = df["score"].apply(
    lambda x: "優" if x >= 90 else ("良" if x >= 80 else "普")
)

# 排序
df_sorted = df.sort_values("score", ascending=False)

import matplotlib.pyplot as plt
import numpy as np

epochs = range(1, 11)
train_loss = [0.95, 0.75, 0.60, 0.48, 0.38,
              0.30, 0.24, 0.19, 0.15, 0.12]
val_loss   = [0.98, 0.80, 0.67, 0.55, 0.46,
              0.39, 0.34, 0.30, 0.27, 0.25]

# 折線圖 — 訓練曲線
plt.figure(figsize=(8, 4))
plt.plot(epochs, train_loss, 'b-o', label='Training Loss')
plt.plot(epochs, val_loss, 'r--s', label='Validation Loss')
plt.title('訓練曲線')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend()
plt.grid(True, alpha=0.3)
plt.tight_layout()
plt.savefig('training_curve.png', dpi=150)
plt.show()

# 子圖 — 同時顯示多張圖
fig, axes = plt.subplots(1, 2, figsize=(12, 4))
axes[0].bar(['A', 'B', 'C'], [10, 20, 15])
axes[1].hist(np.random.randn(1000), bins=30)
plt.show()

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, classification_report

# 1. 載入資料
iris = load_iris()
X, y = iris.data, iris.target

# 2. 切分訓練/測試集（8:2）
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

# 3. 標準化（重要！）
scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test  = scaler.transform(X_test)

# 4. 訓練模型
model = LogisticRegression(max_iter=200)
model.fit(X_train, y_train)

# 5. 評估
y_pred = model.predict(X_test)
print(f"準確率：{accuracy_score(y_test, y_pred):.2%}")
print(classification_report(y_test, y_pred,
      target_names=iris.target_names))

# pip install anthropic
import anthropic

client = anthropic.Anthropic(api_key="YOUR_API_KEY")

# 基本對話
message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[
        {"role": "user",
         "content": "用一句話解釋什麼是機器學習"}
    ]
)
print(message.content[0].text)

# 多輪對話
history = []
def chat(user_msg):
    history.append({"role": "user", "content": user_msg})
    resp = client.messages.create(
        model="claude-sonnet-4-6",
        max_tokens=512,
        messages=history
    )
    ai_msg = resp.content[0].text
    history.append({"role": "assistant", "content": ai_msg})
    return ai_msg

print(chat("什麼是 Python？"))
print(chat("給我一個簡單範例"))

import numpy as np
import pandas as pd
from sklearn.preprocessing import (
    StandardScaler, MinMaxScaler, LabelEncoder
)

data = pd.DataFrame({
    "age":    [25, 30, 22, 35],
    "income": [30000, 50000, 25000, 80000],
    "city":   ["台北", "台中", "高雄", "台北"]
})

# 數值標準化（Z-score）
scaler = StandardScaler()
data["age_scaled"] = scaler.fit_transform(data[["age"]])

# 最小最大縮放（0~1）
mms = MinMaxScaler()
data["income_norm"] = mms.fit_transform(data[["income"]])

# 類別編碼
le = LabelEncoder()
data["city_code"] = le.fit_transform(data["city"])
print(le.classes_)  # ['台中' '台北' '高雄']

# One-Hot 編碼
dummies = pd.get_dummies(data["city"], prefix="city")
data = pd.concat([data, dummies], axis=1)
print(data.head())

# 建立虛擬環境
python3 -m venv ai_env

# 啟動（Mac/Linux）
source ai_env/bin/activate

# 啟動（Windows）
ai_env\Scripts\activate

# 安裝 AI 必備套件
pip install numpy pandas matplotlib scikit-learn
pip install jupyter notebook
pip install anthropic  # Claude API
pip install torch torchvision  # PyTorch

# 儲存套件清單
pip freeze > requirements.txt

# 其他人從清單安裝
pip install -r requirements.txt

# 啟動 Jupyter Notebook
jupyter notebook