Quantum computing has emerged in the past few years as a new computing model that could take today’s computers capabilities to a whole new level. All technology-related media had been publishing all small and possible advances in the field. Although this is a fascinating time for the field, the field itself remains a big mystery.
The premise behind quantum computing is that it can be assets in various applications that are considered essential in today’s technological world, from cybersecurity to medical applications to machine learning. The wide scope of applications is one of the main causes for the attention the field has been receiving.
How can quantum help advance the field of data science? What can it offer that classical computers failed to offer?
By now, you’ve probably heard of “Quantum Machine Learning” or QML. But, what is really quantum about it?
This article aims to shed some light on what quantum machine learning is and the possible ways quantum techniques may enhance and improve classic machine learning.
Let’s take a step back and look at machine learning from a wider perspective. Machine learning consists of two things: data and algorithms. So, when we say quantum machine learning, are we refereeing to the data or the algorithms, or both?
That is a fundamental question to ask ourselves, what is quantum about quantum machine learning? To answer this question, let’s consider the figure below.
Quantum machine learning is a term used to cover 4 types of scenarios:
- Quantum-inspired classical algorithms on classical data: such as tensor network and de-quantized recommendation systems algorithms.
- Classical algorithms are applied to quantum data: such as neural network-based quantum States and optimizing pulse sequences.
- Quantum algorithms are applied to classical data: such as quantum optimization algorithms and quantum classification of classical data.
- Quantum algorithms are applied to quantum data: such as quantum signal processing and quantum hardware modeling.
There are different theories on how can quantum better machine learning. Here are the top 3 arguments:
1. If quantum computers have speedups in linear algebra subroutines, it can speed up machine learning.
We all know that linear algebra is the core of machine learning. In particular, a group of linear algebra applications called BLAS (Basic Linear Algebra Subroutines) is the fundamentals of all machine learning algorithms. These subroutines include matrix multiplication, Fourier transforms, and solving linear systems.
All these subroutines do obtain exponential speedups when ran on a quantum computer. However, when addressing these speedups, what is not mentioned is, to obtain these speedups, we have to have a quantum memory the holds quantum data and communicates with a quantum processor. Then, and only then, we can reach exponential speedups.
Unfortunately, this is not the case. In fact, quantum memories are one of the complex research topics currently, with no concrete version of when/ if a quantum memory can be built.
Does that mean we can’t achieve speedups at all?
Currently, our systems are not pure quantum; our data is classical and is stored in a classical memory. This data is then communicated to a quantum processor. The communication between classical memory and the quantum processor is why an exponential speedup can’t be reached.
Based on the memory and the nature of the linear algebra application used, we can achieve some sort of speedup over the pure classical approaches.
2. Quantum parallelism can help train models more faster
One of the main power sources of quantum computers is their ability to perform quantum superposition. Which enables us to work on various quantum states at the same time.
So, the argument here is, if we can train a model in a state of superposition of all possible training sets, then maybe the training process will be faster and more efficient.
Efficient here can mean one of two things:
- Exponentially fewer data needed to train the model -> which researchers have found is inaccurate. However, some linear speedups may be possible in some cases.
- Train models faster -> this claim follows the speedup resulted from quantizing any classical algorithm following Grover’s algorithm. The result is speedups up to quadratic at best and not exponential.
If I tried running a classical machine learning algorithm on quantum computing, the best I can aim up for is quadratic speedups. If I need more speedups, then the algorithm needs to change as well.
3. Quantum Computers can model highly correlated distributions in a way classical computers can’t.
This is true, 100%. However, while it is correct, recent research results proved that this is insufficient for any quantum advantage. Moreover, it showed that some classical models could outperform quantum ones, even on datasets generated quantumly.
In short, YES. But, we shouldn’t expect exponential speedups at any point soon. Maybe once fully functional quantum computers are built, we can revisit those above arguments and test their validity again.
For now, pursuing quantum machine learning should be for discovering fundamentally new algorithms that could help generate better problems rather than aim for speeding up existing machine learning algorithms.
The way I think of it is, quantum teaches us a new way of looking at things instead of a new angle of looking at the same thing.
Quantum computing has many potentials in improving and reshaping the way technology is at the moment. Even though quantum computing is not really a new field at all — it’s as old as computing itself — it gained traction and much attention in recent years.
Recently, however, quantum has moved from pure theory to practice. That is the reason for the sudden blow up in attention from both media and research. For the first time in quantum history, anyone can access quantum machines and run small to moderate programs on them.
One of the fields that show some promise for improvements caused by quantum techniques is machine learning. But how can it actually help is something that is not very obvious.
In this article, I went through the possible ways quantum can improve machine learning and whether these ways are valid and realistic.
Maybe quantum computing is not that powerful yet, and aren’t currently capable of running machine learning applications. Still, the future is bright and will hold up many surprises and advancements to many more fields, not just machine learning.