Data Science News

A paper co-authored by Matthias C. Caro has been featured on the cover of PRX Quantum. PRX Quantum is a premier journal for quantum information science and technology research. The work was a collaboration with Haimeng Zhao (Caltech & Tsinghua), Laura Lewis (Caltech & Google), Ishaan Kannan (Caltech), Yihui Quek (Harvard & MIT) and Hsin-Yuan Huang (Caltech, Google & MIT).

Characterizing a quantum system by learning its state or unitary evolution is a key tool in developing quantum devices, with applications in practical quantum machine learning, benchmarking, and error mitigation. However, in general, this task requires exponentially many resources. Prior knowledge is required to circumvent this exponential bottleneck. The paper pinpoints the complexity for learning states and unitaries that can be implemented by quantum circuits with a bounded number of gates, a broad setting that is topical for current quantum technologies. When measuring efficiency with respect to the number of accesses to the unknown quantum state or unitary, the paper presents and implements algorithms that are provably optimally efficient. Thereby, this work establishes the equivalence between the complexity of learning quantum states or unitaries and the complexity of creating them. However, it also shows that the data processing necessarily requires exponential computation time under reasonable cryptographic assumptions.

In a paper recently published in Nature Computational Science, led by Prof. Jianfeng Feng, they developed the platform of Digital Twin Brain (DTB) for simulating spiking neuronal networks at the large neuron scale of the human brain on the basis of personalized magnetic resonance imaging data and biological constraints.

A recent publication in Cell, leading by Prof. Jianfeng Feng, presents a comprehensive atlas of proteome-phenome associations (https://proteome-phenome-atlas.com/Link opens in a new window) by systematically mapping 2,920 plasma proteins to the presence and onset of 720 diseases and 986 health-related traits in 53,026 individuals. This atlas provides insights into shared and characteristic biological mechanisms among diseases. The proteomic profiles, coupled with machine learning, identify useful biomarkers and prediction models for multiple health conditions simultaneously. Through integrating protein quantitative trait locus (pQTL) data, this work also illustrates the use of the atlas for causal protein discovery and further drug target prioritization. The proteome-phenome atlas furnishes an extensive resource supporting future research in screening, diagnosis, and treatment of human diseases and advancing precision medicine.

See here https://statsupai.org/quarto_web/site/posts/S11_interview.html for a more detailed discussion on the article.