Sonika Johri

A cheat code for improving your variational quantum algorithms is to study quantum optimal control papers from the last couple of decades

Do reach out to us at Coherent Computing coherentcomputing.com. if you’d like us to help evaluate whether your dataset is a good fit for quantum.

If you’re considering a quantum machine learning project, don’t go in blind: knowing the qubit requirement up front is essential to avoid wasted effort and ensure the project is truly worth pursuing.

These results aren’t just academic - we’ve developed a practical framework within Red Cedar, our QML platform, that you can apply directly to your own datasets.

But they are still well within the reach of the quantum computers that are expected to be viable in the next 5 years! These results support that some of the largest, most information-rich problems in biology and beyond may be prime candidates for quantum machine learning.

However, when applied to subsets of the Tahoe-100M dataset from Tahoe Therapeutics, a transcriptomic dataset with 100 million samples, the required qubits quickly exceed the practical limit for classical simulation.

Within this framework, we find that many medium-sized datasets require only about 20 qubits to encode, and so they aren’t great candidates for quantum learning.

It also allows us to predict how many logical qubits are needed for a quantum model to train to a desired accuracy on a dataset of interest. This is the first encoding framework that connects datasets directly to quantum hardware requirements.

In our paper “How many qubits does a machine learning problem require?” arxiv.org/pdf/2508.20992, we show that bit-bit encoding, a recently developed classical -> quantum encoding technique, makes it possible to encode datasets efficiently into quantum computers without compromising on expressivity.

At Coherent Computing, we’ve been thinking hard about these challenges, and we have a solution.

Or you went with amplitude encoding, and suddenly the cost of loading the data eats up your entire quantum budget unless you compromise on the loading quality. Or perhaps you turned to hybrid quantum-classical models, only to wonder whether the quantum part is even pulling its weight?

You’ve probably wrestled with how to encode it into a quantum computer. Maybe you tried angle encoding with data reuploading - only to realize that boosting expressivity isn’t straightforward.

Do you have a dataset you think could benefit from quantum learning?

Keep reading!

Late to post, but still worth a read: our conceptual paper on a quantum mulitomics platform: arxiv.org/pdf/2506.14080

This wacky SCOTUS opinion could be used against teaching any kind of science, history, or critical reasoning more generally, which might pose "a very real threat of undermining" religious beliefs. Even if it doesn't actually, there's the threat!

On Wednesday I'm giving a talk at the Quantum Mechanics for Modeling Classical Dynamics symposium at the Society for Industrial and Applied Mathematics Conference on Dynamical Systems (DS25) in Denver. Looking forward to the discussions and let's connect if you're around and interested in the topic!

This is a PSA: As far as variational quantum algorithms go, you should approach with skepticism the practical implications of any theoretical result that purports to make a general statement but assumes Haar randomness in the proof :)

What is this madness??

A Practical Framework for Assessing the Performance of Observable Estimation in Quantum Simulation Simulating dynamics of physical systems is a key application of quantum computing, with potential impact in fields such as condensed matter physics and quantum chemistry. However, current quantum algo...

We are proud to have contributed to the latest enhancements to the open-source QED-C Application-Oriented Benchmark suite. Check out this new paper on benchmarking algorithmic techniques for the computation of observables in Hamiltonian simulation.

arxiv.org/abs/2504.09813

I really enjoyed this wide-ranging discussion about quantum physics and the possibilities of using it in computing with Amit Prakash and Dheeraj Pandey! Perfect timing for it to be released on World Quantum Day yesterday :)

youtube.com/watch?v=ysPY...

Please come to Vancouver too!

Bit-bit encoding, optimizer-free training and sub-net initialization: techniques for scalable quantum machine learning Quantum machine learning for classical data is currently perceived to have a scalability problem due to (i) a bottleneck at the point of loading data into quantum states, (ii) the lack of clarity arou...

If you missed the talk, the paper is at arxiv.org/abs/2501.02148

And yesterday I took a break from the meeting to head to Caltech to meet my PhD advisor Ravin Bhatt was was visiting

Very gratifying to present today at the APS meeting in Anaheim California

Come listen and get in touch if you want to discuss either of the topics!

Quantum Computing: Algorithms, Architectures, and Applications 6:00 pm – 8:00 pm, Monday March 17, Session VIR-C01, Virtual-Only, Virtual Room 1

- A Comprehensive Cross-Model Framework for Benchmarking the Performance of Quantum Hamiltonian Simulations, presented by Avimita Chatterjee
summit.aps.org/events/VIR-C...

Quantum Machine Learning: Theory and Training 11:30 am – 2:30 pm, Friday March 21, Session MAR-X34, Anaheim Convention Center, 256A (Level 2)

Two talks with contributions from Coherent Computing at the APS Global Physics Summit this week:

- Techniques for design and training of large quantum machine learning models, presented by me
summit.aps.org/events/MAR-X...

Not commenting on the video but which QML approaches require measuring an exponential number of amplitudes?? For classification, say, the goal would be to read out the assigned class.

Data is often loadable in short depth: Quantum circuits from tensor networks for finance, images, fluids, and proteins Though there has been substantial progress in developing quantum algorithms to study classical datasets, the cost of simply \textit{loading} classical data is an obstacle to quantum advantage. When th...

True if you want to load to very high precision but have you seen works such as
arxiv.org/abs/2309.13108

arxiv.org/abs/2310.05897

To me, it's also a strong indicator that distinct computational paradigms such as those provided by quantum computers will have a role to play in the future of AI.