問題
After a neuron adds up its weighted inputs, an activation function decides its output. Without one, a neural network could only learn straight lines.
Implement activate(name, values) and apply the named function to every value:
| name | formula |
|---|---|
"sigmoid" | 1 / (1 + e^(-x)) |
"relu" | max(0, x) |
"leaky_relu" | x if x > 0, otherwise 0.01 × x |
"tanh" | (e^x - e^(-x)) / (e^x + e^(-x)) |
Input
name: One of "sigmoid", "relu", "leaky_relu" or "tanh".
values: A list of numbers.
Output
Return a list with the activation applied to each value, in the same order.
制約
1 ≤ length of values ≤ 1,000
-50 ≤ each value ≤ 50
例
入力：
name = "sigmoid"
values = [0.0, 2.0, -2.0]
期待される出力：
expected = [0.5, 0.880797, 0.119203]入力：
std::string name = "sigmoid";
std::vector<double> values = {0.0, 2.0, -2.0};
期待される出力：
std::vector<double> expected = {0.5, 0.880797, 0.119203};入力：
string name = "sigmoid";
double[] values = new double[] {0.0, 2.0, -2.0};
期待される出力：
double[] expected = new double[] {0.5, 0.880797, 0.119203};解説
sigmoid(0) = 1 / (1 + 1) = 0.5, sigmoid(2) ≈ 0.880797 and sigmoid(-2) ≈ 0.119203. Sigmoid squeezes any number into the range 0 to 1.
入力：
name = "relu"
values = [-1.5, 0.0, 3.0]
期待される出力：
expected = [0.0, 0.0, 3.0]入力：
std::string name = "relu";
std::vector<double> values = {-1.5, 0.0, 3.0};
期待される出力：
std::vector<double> expected = {0.0, 0.0, 3.0};入力：
string name = "relu";
double[] values = new double[] {-1.5, 0.0, 3.0};
期待される出力：
double[] expected = new double[] {0.0, 0.0, 3.0};解説
ReLU keeps positive values and turns negative values into 0: [0, 0, 3].
解き方のヒント
まずは自分で考えてから、ヒントを 1 つずつ開いてみましょう。
ヒント 1
Each activation is applied independently to every number, so focus on one value at a time.
ヒント 2
Use the formula matching the given name and build a new list from the results.
ヒント 3
Be careful with negative inputs, especially for ReLU and the leaky slope on the nonpositive side.
目安の計算量
O(n) time, O(n) extra space
参考資料
この課題はCodeUtility用に調整されています。上記の説明と制約に従ってください。