Machine learning · federated learning · secure data analytics
Finding structure in data that resists it.
Dr. Mahwish Yousaf builds methods that recover the shape hidden inside high-dimensional measurements, reconstruct readings that were never recorded, and let sensitive data be learned from without being surrendered.
- Field
- Machine learning researcher
- Based in
- Hefei, China
- Doctorate
- USTC, 2021
Citation record
From Google Scholar, checked 2 August 2026. Verify
Research programme
Three problems, one concern.
Each strand asks the same question from a different direction: how do you draw a reliable conclusion from data that is too large, too incomplete, or too sensitive to work with directly?
Making manifold learning usable
Faster, cleaner, and more honest recovery of the shape hidden in high-dimensional data.
Read the detail 02Recovering data that was never recorded
Reconstructing missing traffic measurements from the structure that survives around them.
Read the detail 03Learning without surrendering the data
Cryptography and federated methods applied to settings where the records cannot be pooled.
Read the detailResearch, made visible
Finding the shape inside the noise
Measurements arrive with far more dimensions than the process behind them. Watch scattered samples settle onto the one-dimensional curve they actually came from, with neighbourhood links appearing as the geometry resolves.
Selected work
One paper from each strand.
Titles link to the version of record. Restricted publisher files are not rehosted here.
MNT-TNN: Spatiotemporal Traffic Data Imputation via Compact Multimode Nonlinear Transform-Based Tensor Nuclear Norm
Traffic sensors fail, and the gaps they leave distort everything computed downstream. This work recovers the missing readings by treating the network as a tensor over space and time, and modelling the structure that survives across all three modes at once.
Read via DOIAn Extended-Isomap for High-Dimensional Data Accuracy and Efficiency: A Comprehensive Survey
A map of the field: what Isomap gets wrong, which of the many proposed repairs actually hold up, and where the open problems still are.
Read via DOIElliptic Curve Cryptography: Applications, Challenges, Recent Advances, and Future Trends — A Comprehensive Survey
A wide survey of where elliptic-curve cryptography is actually deployed, what breaks it, and which advances have held up in practice.
Read via DOIVerified identity
Check the record independently.
Every profile below was confirmed against her institution, email, or publication record before it was linked.
Open to conversation
Collaboration, supervision, and invited talks.
For joint work, research appointments, student enquiries, or speaking invitations, write directly.
Get in touch