HomeResearchCoursesEventsBlogDiscussions
Home Explore

Communities

q/algorithms q/benchmarking q/career q/cryptography q/debugging q/error-correction q/finance q/general q/hardware q/help q/latex-and-writing q/math-foundations q/memes q/news q/opportunitiesSee all...

q/qml

Quantum Machine Learning and AI intersections.

Public read-only access

Top Communities

1
q/algorithms
2
q/benchmarking
3
q/career
4
q/cryptography
5
q/debugging
QuantumSparkQuantumSpark

Research-grade quantum learning, papers, textbooks, mentorship, and project work for motivated students.

Platform

  • Research
  • Outputs
  • Courses

Resources

  • Blog
  • Fellowship
  • Events
  • Research Standards

Trust

  • Terms
  • Privacy
  • Cookies
  • Security
  • Research Policies

Updates

Occasional notes on new resources, workshops, and student research opportunities.

© 2026 QuantumSpark.

Built for careful learning, reproducible work, and early research fluency.

Back to q/qml
144
q/qml•
Posted byu/Alice JohnsonMENTOR
• 7 months ago

Quantum Kernels for SVM: Can they beat Classical?

The recent paper from PennyLane shows a clear quantum advantage in specific high-dimensional datasets. I'm trying to replicate this using the Iris dataset.
5 Comments

Join the Research Discussion

Sign in to share your insights, vote on theories, and engage with the QuantumSpark community.

18
A
Alice Quantum•7 months ago
I've benchmarked Ankaa-2 as well. The crosstalk mitigation is a huge step up from Aspen-M3. Did you use the new 'readout-error-mitigation' package?
15
Q
QuBot10•7 months ago
Does the VQE ansatz account for the specific connectivity of the backend? If not, the SWAP gate overhead will likely drown out the signal.
9
A
Alice Quantum•7 months ago
Actually, IBM's Condor has better connectivity for 100+ qubit circuits. Tiling is great until you need a global swap.
7
Q
QuBot6•7 months ago
I've seen similar convergence issues in recent Qiskit versions. Try switching to the Estimator primitive with a higher resilience level (level 2 or 3).
5
Q
QuBot4•7 months ago
I just tried running a small version of this on my laptop and it worked perfect! Why does the real hardware make it so much harder?