Machine Learning Foundations · Quiz 1 · May 2024

**Answer: 3** **Topic:** Linear Algebra — *Eigenvalues from Known Eigenvectors* ### Step-by-Step Solution **Step 1:** We are given that $v_1 = \begin{pmatrix} 1 \\ 0 \\ -1 \end{pmatrix}$ is an eigenvector of $A = \begin{pmatrix} 2 & 1 & 1 \\ a & 3 & 2 \\ 3 & b & c \end{pmatrix}$ with eigenvalue $\alpha$, which means $A v_1 = \alpha v_1$. **Step 2:** Compare the first component (Row 1) of $A v_1 = \alpha v_1$: $$2(1) + 1(0) + 1(-1) = \alpha(1) \implies 2 - 1 = \alpha \implies \alpha = 1$$ **Step 3:** Similarly, $v_2 = \begin{pmatrix} 0 \\ 1 \\ -1 \end{pmatrix}$ is an eigenvector of $A$ with eigenvalue $\beta$, meaning $A v_2 = \beta v_2$. Compare the second component (Row 2) of $A v_2 = \beta v_2$: $$a(0) + 3(1) + 2(-1) = \beta(1) \implies 3 - 2 = \beta \implies \beta = 1$$ **Step 4:** Compute the requested linear combination $\alpha + 2\beta$: $$\alpha + 2\beta = 1 + 2(1) = 3$$ **Key Takeaway / Tip:** Because $v_1$ has a $0$ in its second component, $a$ disappears from Row 2 of $A v_2$, allowing us to find both $\alpha = 1$ and $\beta = 1$ before even solving for $a, b, c$!

**Answer: A, C, D** **Topic:** Linear Algebra — *Solving for Unknown Matrix Entries from Eigenvectors* ### Step-by-Step Solution **Step 1:** Use $A v_1 = 1 \cdot v_1$ with $v_1 = (1, 0, -1)^T$ to find $a$ and $c$: - Row 2: $a(1) + 3(0) + 2(-1) = 0 \implies a - 2 = 0 \implies a = 2$ - Row 3: $3(1) + b(0) + c(-1) = -1 \implies 3 - c = -1 \implies c = 4$ **Step 2:** Use $A v_2 = 1 \cdot v_2$ with $v_2 = (0, 1, -1)^T$ and $c = 4$ to find $b$: - Row 3: $3(0) + b(1) + 4(-1) = -1 \implies b - 4 = -1 \implies b = 3$ Thus, $a + b + c = 2 + 3 + 4 = 9$ (**D is true**). **Step 3:** Check the remaining statements: - **A is true**: $v_1 = (1, 0, -1)^T$ and $v_2 = (0, 1, -1)^T$ are not scalar multiples of each other, so they are linearly independent. - **B is false**: $v_1^T v_2 = (1)(0) + (0)(1) + (-1)(-1) = 1 \ne 0$, so $v_1$ and $v_2$ are not orthogonal. - **C is true**: Multiply $A = \begin{pmatrix} 2 & 1 & 1 \\ 2 & 3 & 2 \\ 3 & 3 & 4 \end{pmatrix}$ by $v_3 = \begin{pmatrix} 1 \\ 2 \\ 3 \end{pmatrix}$: $$A v_3 = \begin{pmatrix} 2(1) + 1(2) + 1(3) \\ 2(1) + 3(2) + 2(3) \\ 3(1) + 3(2) + 4(3) \end{pmatrix} = \begin{pmatrix} 7 \\ 14 \\ 21 \end{pmatrix} = 7 \begin{pmatrix} 1 \\ 2 \\ 3 \end{pmatrix}$$ Thus, $(1, 2, 3)^T$ is the third eigenvector of $A$ (with eigenvalue $\lambda_3 = 7$). **Step 4:** Therefore, **A, C, and D** are correct. **Key Takeaway / Tip:** You can double-check the third eigenvalue using the trace: $\text{tr}(A) = 2 + 3 + 4 = 9 = \alpha + \beta + \lambda_3 = 1 + 1 + 7$. A: **Correct:** Option A (

) matches our derived result under Solving for Unknown Matrix Entries from Eigenvectors. You can double-check the third eigenvalue using the trace: $\text{tr}(A) = 2 + 3 + 4 = 9 = \alpha + \beta + \lambda_3 = 1 + 1 + 7$. B: **Incorrect:** Option B (

) does not satisfy the conditions derived in the step-by-step analysis for Solving for Unknown Matrix Entries from Eigenvectors (correct choice: A, C, D). C: **Correct:** Option C (

) matches our derived result under Solving for Unknown Matrix Entries from Eigenvectors. You can double-check the third eigenvalue using the trace: $\text{tr}(A) = 2 + 3 + 4 = 9 = \alpha + \beta + \lambda_3 = 1 + 1 + 7$. D: **Correct:** Option D (

) matches our derived result under Solving for Unknown Matrix Entries from Eigenvectors. You can double-check the third eigenvalue using the trace: $\text{tr}(A) = 2 + 3 + 4 = 9 = \alpha + \beta + \lambda_3 = 1 + 1 + 7$.

**Answer: A, C, D** **Topic:** Linear Algebra — *Eigenvalues and Diagonalizability of a Symmetric Matrix* ### Step-by-Step Solution **Step 1:** Inspect $A = \begin{pmatrix} 0 & 1 & 1 \\ 1 & 0 & 1 \\ 1 & 1 & 0 \end{pmatrix}$: - Since $A^T = A$, **$A$ is symmetric** (**A is true**). - By the Real Spectral Theorem, every real symmetric matrix is **orthogonally diagonalizable** (**C is true**). **Step 2:** Find the eigenvalues of $A$: notice that $A = J - I$, where $J = \mathbf{1}\mathbf{1}^T$ is the $3 \times 3$ all-ones matrix. Since $J$ is rank-1 with eigenvalues $3, 0, 0$, subtracting $I$ shifts every eigenvalue by $-1$: $$\lambda_1 = 3 - 1 = 2, \quad \lambda_2 = 0 - 1 = -1, \quad \lambda_3 = 0 - 1 = -1$$ Thus, **the eigenvalues of $A$ are $-1, -1, 2$** (**D is true**). **Step 3:** Since none of the eigenvalues is $0$, $\det(A) = (-1)(-1)(2) = 2 \ne 0$, so $A$ is invertible (**B is false**). Therefore, **A, C, and D** are correct. **Key Takeaway / Tip:** Writing $A = \mathbf{1}\mathbf{1}^T - I$ gives the eigenvalues of $A$ immediately without expanding a cubic characteristic polynomial. A: **Correct:** Option A (A is a symmetric matrix.) matches our derived result under Eigenvalues and Diagonalizability of a Symmetric Matrix. Writing $A = \mathbf{1}\mathbf{1}^T - I$ gives the eigenvalues of $A$ immediately without expanding a cubic characteristic polynomial. B: **Incorrect:** Option B (A is a diagonal matrix.) does not satisfy the conditions derived in the step-by-step analysis for Eigenvalues and Diagonalizability of a Symmetric Matrix (correct choice: A, C, D). C: **Correct:** Option C (A is orthogonally diagonalizable.) matches our derived result under Eigenvalues and Diagonalizability of a Symmetric Matrix. Writing $A = \mathbf{1}\mathbf{1}^T - I$ gives the eigenvalues of $A$ immediately without expanding a cubic characteristic polynomial. D: **Correct:** Option D (Eigenvalues of A are -1, -1 and 2.) matches our derived result under Eigenvalues and Diagonalizability of a Symmetric Matrix. Writing $A = \mathbf{1}\mathbf{1}^T - I$ gives the eigenvalues of $A$ immediately without expanding a cubic characteristic polynomial.

**Answer: A, C** **Topic:** Linear Algebra — *Column Space and Null Space of a Full-Column-Rank Matrix* ### Step-by-Step Solution **Step 1:** Let $B = [v_1 \; v_2] \in \mathbb{R}^{3 \times 2}$ be the $3 \times 2$ matrix whose columns are two linearly independent vectors $v_1, v_2 \in \mathbb{R}^3$. Thus, $\text{rank}(B) = 2$. **Step 2:** Analyze the **Null Space** $\mathcal{N}(B) = \{x \in \mathbb{R}^2 : Bx = 0\}$: - By the Rank-Nullity Theorem, $\dim(\mathcal{N}(B)) = n - \text{rank}(B) = 2 - 2 = 0$. - Equivalently, $Bx = x_1 v_1 + x_2 v_2 = 0 \implies x_1 = x_2 = 0$ by linear independence of $v_1, v_2$. Thus, **the null space of $B$ is the zero space $\{0\}$** (**A is true**, **B is false**). **Step 3:** Analyze the **Column Space** $\mathcal{C}(B) = \text{span}(v_1, v_2) \subset \mathbb{R}^3$: - Because $v_1$ and $v_2$ are two linearly independent vectors in $\mathbb{R}^3$, their span is a 2-dimensional subspace of $\mathbb{R}^3$, which geometrically is **a plane passing through the origin in $\mathbb{R}^3$** (**C is true**, **D is false**). **Step 4:** Therefore, **A and C** are correct. **Key Takeaway / Tip:** The null space of an $m \times n$ matrix lives in $\mathbb{R}^n$ (here $\mathbb{R}^2$), whereas the column space lives in $\mathbb{R}^m$ (here $\mathbb{R}^3$). A: **Correct:** Option A (

) matches our derived result under Column Space and Null Space of a Full-Column-Rank Matrix. The null space of an $m \times n$ matrix lives in $\mathbb{R}^n$ (here $\mathbb{R}^2$), whereas the column space lives in $\mathbb{R}^m$ (here $\mathbb{R}^3$). B: **Incorrect:** Option B (

) does not satisfy the conditions derived in the step-by-step analysis for Column Space and Null Space of a Full-Column-Rank Matrix (correct choice: A, C). C: **Correct:** Option C (

) matches our derived result under Column Space and Null Space of a Full-Column-Rank Matrix. The null space of an $m \times n$ matrix lives in $\mathbb{R}^n$ (here $\mathbb{R}^2$), whereas the column space lives in $\mathbb{R}^m$ (here $\mathbb{R}^3$). D: **Incorrect:** Option D (

) does not satisfy the conditions derived in the step-by-step analysis for Column Space and Null Space of a Full-Column-Rank Matrix (correct choice: A, C).

**Answer: A, B, C** **Topic:** Linear Algebra — *Properties of a Rank-1 Orthogonal Projection Matrix* ### Step-by-Step Solution **Step 1:** For $a = \begin{pmatrix} 1 \\ 2 \end{pmatrix}$, the matrix that projects any vector in $\mathbb{R}^2$ orthogonally onto the line spanned by $a$ is: $$P = \frac{a a^T}{a^T a} = \frac{1}{1^2 + 2^2} \begin{pmatrix} 1 \\ 2 \end{pmatrix} \begin{pmatrix} 1 & 2 \end{pmatrix} = \frac{1}{5} \begin{pmatrix} 1 & 2 \\ 2 & 4 \end{pmatrix}$$ **Step 2:** Verify each statement: - **A is true**: $P^T = P$ (every orthogonal projection matrix is symmetric), so $P$ is orthogonally diagonalizable. - **B is true**: Every projection matrix is idempotent ($P^2 = P$), which implies $P^3 = P$ and $P^4 = P$. - **C is true**: The projection of $b = \begin{pmatrix} 1 \\ -1 \end{pmatrix}$ onto $a$ is: $$P b = \frac{a^T b}{a^T a} a = \frac{1(1) + 2(-1)}{5} a = -\frac{1}{5} a$$ - **D is false**: Because $P$ projects all of $\mathbb{R}^2$ onto a 1-dimensional line $\text{span}(a)$, $\text{rank}(P) = 1 < 2$, so $P$ is singular (non-invertible). **Step 3:** Therefore, **A, B, and C** are correct. **Key Takeaway / Tip:** Any projection onto a proper subspace has $\text{rank}(P) = \dim(\text{subspace}) < n$ and is therefore never invertible. A: **Correct:** Option A (

) matches our derived result under Properties of a Rank-1 Orthogonal Projection Matrix. Any projection onto a proper subspace has $\text{rank}(P) = \dim(\text{subspace}) < n$ and is therefore never invertible. B: **Correct:** Option B (

) matches our derived result under Properties of a Rank-1 Orthogonal Projection Matrix. Any projection onto a proper subspace has $\text{rank}(P) = \dim(\text{subspace}) < n$ and is therefore never invertible. C: **Correct:** Option C (

) matches our derived result under Properties of a Rank-1 Orthogonal Projection Matrix. Any projection onto a proper subspace has $\text{rank}(P) = \dim(\text{subspace}) < n$ and is therefore never invertible. D: **Incorrect:** Option D (

) does not satisfy the conditions derived in the step-by-step analysis for Properties of a Rank-1 Orthogonal Projection Matrix (correct choice: A, B, C).

**Answer: A** **Topic:** Linear Regression — *Ordinary Least Squares Simple Linear Regression* ### Step-by-Step Solution **Step 1:** Given the $n = 4$ data points $(0, 1), (2, 2), (3, 4), (5, 3)$, compute the sample means $\bar{x}$ and $\bar{y}$: $$\bar{x} = \frac{0 + 2 + 3 + 5}{4} = 2.5, \quad \bar{y} = \frac{1 + 2 + 4 + 3}{4} = 2.5$$ **Step 2:** Compute $S_{xx} = \sum_{i=1}^4 (x_i - \bar{x})^2$ and $S_{xy} = \sum_{i=1}^4 (x_i - \bar{x})(y_i - \bar{y})$: - $(x - \bar{x}) = [-2.5, \; -0.5, \; 0.5, \; 2.5] \implies S_{xx} = 6.25 + 0.25 + 0.25 + 6.25 = 13$ - $(y - \bar{y}) = [-1.5, \; -0.5, \; 1.5, \; 0.5] \implies S_{xy} = (-2.5)(-1.5) + (-0.5)(-0.5) + (0.5)(1.5) + (2.5)(0.5) = 3.75 + 0.25 + 0.75 + 1.25 = 6$ **Step 3:** Calculate the least-squares slope $b_1$ and intercept $b_0$: $$b_1 = \frac{S_{xy}}{S_{xx}} = \frac{6}{13} \approx 0.4615 \approx 0.46$$ $$b_0 = \bar{y} - b_1 \bar{x} = 2.5 - \frac{6}{13}(2.5) = \frac{32.5 - 15}{13} = \frac{17.5}{13} \approx 1.346 \approx 1.34$$ **Step 4:** Thus, the best-fitting line is $\hat{y} = 1.34 + 0.46x$, which matches **A**. **Key Takeaway / Tip:** Always check that your fitted regression line passes through the centroid $(\bar{x}, \bar{y}) = (2.5, 2.5)$: $1.346 + 0.4615(2.5) = 2.5$. A: **Correct:** Option A (

) matches our derived result under Ordinary Least Squares Simple Linear Regression. Always check that your fitted regression line passes through the centroid $(\bar{x}, \bar{y}) = (2.5, 2.5)$: $1.346 + 0.4615(2.5) = 2.5$. B: **Incorrect:** Option B (

) does not satisfy the conditions derived in the step-by-step analysis for Ordinary Least Squares Simple Linear Regression (correct choice: A). C: **Incorrect:** Option C (

) does not satisfy the conditions derived in the step-by-step analysis for Ordinary Least Squares Simple Linear Regression (correct choice: A). D: **Incorrect:** Option D (

) does not satisfy the conditions derived in the step-by-step analysis for Ordinary Least Squares Simple Linear Regression (correct choice: A).

**Answer: B** **Topic:** Calculus — *Differentiability of a Piecewise Function* ### Step-by-Step Solution **Step 1:** For the piecewise function $f(x) = \begin{cases} a x^2 + b x + 1, & x \le 0 \\ c x + d, & x > 0 \end{cases}$ to be differentiable at $x = 0$, it must first be **continuous** at $x = 0$: $$\lim_{x \to 0^-} f(x) = \lim_{x \to 0^+} f(x) \implies a(0)^2 + b(0) + 1 = c(0) + d \implies d = 1$$ **Step 2:** Next, the left-hand derivative and right-hand derivative at $x = 0$ must be equal: $$f'_-(0) = \left.\frac{d}{dx}(a x^2 + b x + 1)\right|_{x=0} = 2a(0) + b = b$$ $$f'_+(0) = \left.\frac{d}{dx}(c x + d)\right|_{x=0} = c$$ Equating the one-sided derivatives gives $b = c$ (while $a$ can be any real number). **Step 3:** Thus, the required conditions are $b = c$ and $d = 1$, which matches **B**. **Key Takeaway / Tip:** Differentiability at a boundary point requires both continuity (matching function values) and smoothness (matching left and right derivatives). A: **Incorrect:** Option A (

) does not satisfy the conditions derived in the step-by-step analysis for Differentiability of a Piecewise Function (correct choice: B). B: **Correct:** Option B (

) matches our derived result under Differentiability of a Piecewise Function. Differentiability at a boundary point requires both continuity (matching function values) and smoothness (matching left and right derivatives). C: **Incorrect:** Option C (

) does not satisfy the conditions derived in the step-by-step analysis for Differentiability of a Piecewise Function (correct choice: B). D: **Incorrect:** Option D (

) does not satisfy the conditions derived in the step-by-step analysis for Differentiability of a Piecewise Function (correct choice: B).

**Answer: A** **Topic:** Multivariate Calculus — *Differentiability and Directional Derivatives* ### Step-by-Step Solution **Step 1:** Recall the fundamental theorem of multivariable differentiability: if a function $f(x, y)$ is **differentiable** at a point $(a, b)$, then: 1. $f$ is continuous at $(a, b)$. 2. The directional derivative $D_u f(a, b) = \nabla f(a, b) \cdot u$ **exists in every direction** $u$ (including the partial derivatives along the coordinate axes). **Step 2:** Compare with the given options: - **A**: 'The directional derivative exists in every direction at $(a, b)$' — **True**. - **B**: False, differentiability guarantees continuity. - **C**: False, partial derivatives must exist when $f$ is differentiable. - **D**: False, $(a, b)$ only has a zero gradient if it happens to be a stationary point. **Step 3:** Thus, **A** is correct. **Key Takeaway / Tip:** Differentiability at $(a, b)$ implies the existence of all directional derivatives via $D_u f(a, b) = \nabla f(a, b) \cdot u$, though the converse is not generally true. A: **Correct:** Option A (The directional derivative exists in every direction.) matches our derived result under Differentiability and Directional Derivatives. Differentiability at $(a, b)$ implies the existence of all directional derivatives via $D_u f(a, b) = \nabla f(a, b) \cdot u$, though the converse is not generally true. B: **Incorrect:** Option B (The directional derivative is the same in every direction.) does not satisfy the conditions derived in the step-by-step analysis for Differentiability and Directional Derivatives (correct choice: A). C: **Incorrect:** Option C (The directional derivative can only be calculated along the axes.) does not satisfy the conditions derived in the step-by-step analysis for Differentiability and Directional Derivatives (correct choice: A). D: **Incorrect:** Option D (The directional derivative does not exist.) does not satisfy the conditions derived in the step-by-step analysis for Differentiability and Directional Derivatives (correct choice: A).

**Answer: C** **Topic:** Calculus — *Existence of First-Order Taylor Linear Approximation at x = 0* ### Step-by-Step Solution **Step 1:** A function $f(x)$ can be approximated linearly around $x = 0$ via $L(x) = f(0) + f'(0)x$ if and only if $f(0)$ and the derivative $f'(0)$ are both defined and finite. **Step 2:** Check each function at $x = 0$: - **A ($e^x$)**: $f(0) = 1, f'(0) = 1 \implies L(x) = 1 + x$ (valid). - **B ($\sin(x)$)**: $f(0) = 0, f'(0) = \cos(0) = 1 \implies L(x) = x$ (valid). - **C ($\ln(x)$)**: $\ln(x)$ is only defined for $x > 0$; both $\ln(0)$ and $\left.\frac{1}{x}\right|_{x=0}$ are **undefined** (blow up to $-\infty$ and $+\infty$). - **D ($\cos(x)$)**: $f(0) = 1, f'(0) = -\sin(0) = 0 \implies L(x) = 1$ (valid). **Step 3:** Therefore, **$f(x) = \ln(x)$** cannot be linearly approximated around $x = 0$ (**C**). **Key Takeaway / Tip:** While $\ln(1 + x) \approx x$ has a well-known linear approximation around $x = 0$, $\ln(x)$ itself has a vertical asymptote at $x = 0$. A: **Incorrect:** Option A (

) does not satisfy the conditions derived in the step-by-step analysis for Existence of First-Order Taylor Linear Approximation at x = 0 (correct choice: C). B: **Incorrect:** Option B (

) does not satisfy the conditions derived in the step-by-step analysis for Existence of First-Order Taylor Linear Approximation at x = 0 (correct choice: C). C: **Correct:** Option C (

) matches our derived result under Existence of First-Order Taylor Linear Approximation at x = 0. While $\ln(1 + x) \approx x$ has a well-known linear approximation around $x = 0$, $\ln(x)$ itself has a vertical asymptote at $x = 0$. D: **Incorrect:** Option D (

) does not satisfy the conditions derived in the step-by-step analysis for Existence of First-Order Taylor Linear Approximation at x = 0 (correct choice: C).

**Answer: 20.62** **Topic:** Multivariate Calculus — *Magnitude (Euclidean Norm) of the Gradient Vector* ### Step-by-Step Solution **Step 1:** Compute the partial derivatives of $f(x, y) = x^2 y + 3x y^2$: - $\frac{\partial f}{\partial x} = 2xy + 3y^2$ - $\frac{\partial f}{\partial y} = x^2 + 6xy$ **Step 2:** Evaluate the gradient vector $\nabla f(x, y)$ at the point $(1, 2)$: - $\frac{\partial f}{\partial x}(1, 2) = 2(1)(2) + 3(2)^2 = 4 + 12 = 16$ - $\frac{\partial f}{\partial y}(1, 2) = (1)^2 + 6(1)(2) = 1 + 12 = 13$ So $\nabla f(1, 2) = (16, 13)$. **Step 3:** Compute the Euclidean length (norm) of $\nabla f(1, 2)$: $$\|\nabla f(1, 2)\| = \sqrt{16^2 + 13^2} = \sqrt{256 + 169} = \sqrt{425} \approx 20.6155 \approx 20.62$$ **Key Takeaway / Tip:** The length of the gradient vector $\|\nabla f(1, 2)\| = \sqrt{425} \approx 20.62$ also equals the maximum rate of increase of $f$ at $(1, 2)$.

**Answer: 0.2** **Topic:** Classification — *Linear Threshold Classifier Misclassification Rate* ### Step-by-Step Solution **Step 1:** For each of the $n = 5$ applicants, compute the linear score $z = w^T x = 0.6 x_1 + 0.3 x_2 + 0.1 x_3$, apply the step function $\hat{y} = 1$ if $z \ge 3$ else $0$, and compare $\hat{y}$ with the true label $y$: **Step 2:** Evaluate all 5 data points: 1. $x^{(1)} = [1, 5, 0.8], \; y^{(1)} = 1$: $$z^{(1)} = 0.6(1) + 0.3(5) + 0.1(0.8) = 0.6 + 1.5 + 0.08 = 2.18 < 3 \implies \hat{y}^{(1)} = 0 \ne 1\text{ (Misclassified!)}$$ 2. $x^{(2)} = [0, 3, 0.9], \; y^{(2)} = 0$: $$z^{(2)} = 0.6(0) + 0.3(3) + 0.1(0.9) = 0 + 0.9 + 0.09 = 0.99 < 3 \implies \hat{y}^{(2)} = 0 = y^{(2)}\text{ (Correct)}$$ 3. $x^{(3)} = [1, 2, 0.7], \; y^{(3)} = 0$: $$z^{(3)} = 0.6(1) + 0.3(2) + 0.1(0.7) = 0.6 + 0.6 + 0.07 = 1.27 < 3 \implies \hat{y}^{(3)} = 0 = y^{(3)}\text{ (Correct)}$$ 4. $x^{(4)} = [1, 8, 0.6], \; y^{(4)} = 1$: $$z^{(4)} = 0.6(1) + 0.3(8) + 0.1(0.6) = 0.6 + 2.4 + 0.06 = 3.06 \ge 3 \implies \hat{y}^{(4)} = 1 = y^{(4)}\text{ (Correct)}$$ 5. $x^{(5)} = [0, 7, 0.5], \; y^{(5)} = 0$: $$z^{(5)} = 0.6(0) + 0.3(7) + 0.1(0.5) = 0 + 2.1 + 0.05 = 2.15 < 3 \implies \hat{y}^{(5)} = 0 = y^{(5)}\text{ (Correct)}$$ **Step 3:** Compute the fraction of misclassified samples: $$\text{Misclassification Fraction} = \frac{1}{5} = 0.2$$ **Key Takeaway / Tip:** Check whether the question asks for the **fraction** ($1/5 = 0.2$) or the **count** ($1$) of misclassified samples; here it asks for the fraction ($0.2$).

**Answer: C** **Topic:** Machine Learning Fundamentals — *Clustering for Customer Segmentation* ### Step-by-Step Solution **Step 1:** In customer segmentation without pre-existing group labels, the goal is to discover natural homogeneous groups of customers based on similarity in their purchasing behavior and demographics. **Step 2:** This is the canonical definition of **Clustering** (an unsupervised learning task that identifies inherent groupings in unlabeled data). **Step 3:** Compare with the other options: - **A**: Predicting a continuous numerical value is *Regression*. - **B**: Assigning items to predefined labeled categories is *Supervised Classification*. - **C**: 'Identifying inherent groupings in the customer data without predefined labels' — **Correct**. - **D**: Reducing the number of features is *Dimensionality Reduction*. **Step 4:** Thus, **C** is correct. **Key Takeaway / Tip:** Clustering is an unsupervised learning method because the segments/clusters are discovered directly from the data without predefined ground-truth labels. A: **Incorrect:** Option A (Predicting the exact number of sales for each customer segment.) does not satisfy the conditions derived in the step-by-step analysis for Clustering for Customer Segmentation (correct choice: C). B: **Incorrect:** Option B (Classifying customers into segments based on labeled customer data.) does not satisfy the conditions derived in the step-by-step analysis for Clustering for Customer Segmentation (correct choice: C). C: **Correct:** Option C (Identifying inherent groupings in customer data without predefined labels.) matches our derived result under Clustering for Customer Segmentation. Clustering is an unsupervised learning method because the segments/clusters are discovered directly from the data without predefined ground-truth labels. D: **Incorrect:** Option D (Reducing the number of features in the customer dataset.       **Business Analytics** **Section Id :** 64065359212 **Section Number :** 5 **Section type :** Online **Mandatory o...) does not satisfy the conditions derived in the step-by-step analysis for Clustering for Customer Segmentation (correct choice: C).