NPTEL Week 4 – Detailed Solutions
1. 3-qubit phase-flip detection: which error sets are detectable?
Consider the 3-qubit phase-flip detection code, i.e. the repetition code in the X basis. One
convenient stabilizer description is
S = ⟨X1 X2 , X2 X3 ⟩, ZL = Z1 Z2 Z3 , XL = X1 = X2 = X3 on the codespace. (1)
A single-qubit Zj anticommutes with Xj and commutes with Xk for k ̸= j, so the syndrome
bits (s1 , s2 ) measured by (X1 X2 , X2 X3 ) are
Z1 : (1, 0), Z2 : (1, 1), Z3 : (0, 1). (2)
Hence all single-qubit Z errors are mutually distinguishable (and therefore detectable). Now
test each option:
(a) {ZII, IZI, III}: these are {Z1 , Z2 , I}, which are mutually distinguishable (and I is trivially
detectable). Detectable.
(b) {ZII, IZI, ZZI}: syndromes of Z1 and Z2 are distinct as above. For Z1 Z2 , note that
Z1 Z2 commutes with X1 X2 (two anticommutes) but anticommutes with X2 X3 , giving
syndrome (0, 1). This equals the Z3 single-error pattern. Detectability only requires that
the corrupted states {E |ψL ⟩} be mutually orthogonal for distinct E in the set; it does
not require unique identification. Since Z1 Z2 differs from Z1 and Z2 by operators that
anticommute with some stabilizer, one checks ⟨ψL | (Z1 Z2 )† Zj |ψL ⟩ = 0 for j = 1, 2 and any
codespace |ψL ⟩, hence the three error images are mutually orthogonal. Detectable.
(c) {ZII, IZI, IIZ}: all three are distinct single-Z errors; by the table above they are mutually
orthogonal. Detectable.
(d) {ZZI, IZI, ZZZ}: ZZZ = ZL is a logical Z and therefore preserves the codespace; for
a logical state |ψL ⟩, ZZZ |ψL ⟩ remains in the codespace and is not orthogonal to |ψL ⟩
in general, violating detectability against I (and against other errors in the set). Not
detectable as a set.
Correct: (a), (b), (c).
2. Role of the classical optimizer in VQE
In VQE, a parametrized ansatz |ψ(θ)⟩ is prepared on quantum hardware; the classical loop
updates parameters θ to minimize the measured energy E(θ) = ⟨ψ(θ)| H |ψ(θ)⟩. It neither
diagonalizes H (that would be exact classical eigensolving) nor applies gates; it simply proposes
new θ.
Correct: (a)
1
3. Role of the discriminator in a GAN
The discriminator is trained to distinguish real data from generator-produced samples, providing
gradients that drive the generator.
Correct: (c)
4. Bit-flip channel basics
The bit-flip channel with flip probability p is
E(ρ) = (1 − p)ρ + p XρX. (3)
√
Hence the qubit is unchanged with probability 1 − p; it is not 1 − p, and flips with probability
p (not 1).
Correct: (b)
5. Main objective of VQE
VQE variationally estimates the ground-state energy of a many-body Hamiltonian H by mini-
mizing E(θ).
Correct: (d)
6. Nine-qubit Shor code can correct what?
The Shor code concatenates a phase-flip repetition with a bit-flip repetition, enabling correction
of arbitrary single-qubit errors. Any single-qubit error aI + bX + cY + dZ decomposes into Pauli
errors, and the code corrects X and Z (hence also Y = iXZ) on any one qubit. Consequently
both statements (c) and (d) are true.
Correct: (c), (d)
7. 3-qubit bit-flip code: syndrome (s1 , s2 ) = (1, 1)
For the bit-flip repetition code (in the Z basis) with stabilizers Z1 Z2 and Z2 Z3 , the parity checks
detect X-type (bit-flip) errors. A flip on
qubit 1 : (1, 0), (4)
qubit 2 : (1, 1), (5)
qubit 3 : (0, 1). (6)
Therefore (1, 1) points to qubit 2 flipped.
Correct: (b)
8. Smallest QECC correcting arbitrary single-qubit error
The quantum Hamming bound implies n ≥ 5 for a nondegenerate [[n, 1, 3]] code. The [[5, 1, 3]]
five-qubit code saturates this and corrects any single-qubit error, while n = 3 can only handle a
single type (bit- or phase-) of flip and n = 7, 9 are larger.
Correct: (b)
2
9. Objective of the generator in a GAN
The generator aims to produce samples that fool the discriminator (minimax objective), i.e.
make its outputs look real.
Correct: (b)
10. Examples of variational quantum algorithms
VQE and QAOA are canonical VQAs; quantum convolutional neural networks (QCNNs) are
also typically trained variationally. Grover’s algorithm is not variational.
Correct: (a), (c), (d)
