Machine learning with particle physics

[quantknight]

In which fields of physics can we apply machine learning concepts? If I take data science master's with minor in particle physics or astrophysics, what are the possibilities that I would be employed as a physics related data scientist? (I am a programmer and I would like to change my career into employment or research in physics)

 

[StatGuy2000]

First of all, it's important to keep in mind what you mean by "physics related". Are you asking about

1. organizations/companies where you can apply the skills you gained studying particle physics or astrophysics along with your data science masters, or

2. are you looking at places that are directly involved in physics?

If #1, then many of the skill sets involved in both data science and physics can be applied to a wide range of businesses -- lots of insurance companies and market research companies, for example, employ data scientists with a background in physics, in large part because physics graduates often possesses key skillsets (mathematical modelling, programming, analysis of large datasets) that these businesses require.

If you are talking about #2, these may be more limited. Most work involving astrophysics or particle physics are based at a relatively small number of research institutions, either affiliated with universities or with national laboratories (think Fermi labs, CERN in Switzerland, etc.) Only a relatively small number of physics PhD graduates would in general work in these institutions. Statisticians and computer scientists specializing in machine learning have collaborated with physicists in these areas, but my speculation would be that these would be academics in these fields, so I have my doubts about whether a masters in data science would be enough.

 

[quantknight]

If you are talking about #2, these may be more limited. Most work involving astrophysics or particle physics are based at a relatively small number of research institutions, either affiliated with universities or with national laboratories (think Fermi labs, CERN in Switzerland, etc.) Only a relatively small number of physics PhD graduates would in general work in these institutions. Statisticians and computer scientists specializing in machine learning have collaborated with physicists in these areas, but my speculation would be that these would be academics in these fields, so I have my doubts about whether a masters in data science would be enough.

Thanks for the reply statguy. By physics related, I meant the second option you stated. I am already a data mining analyst, so building a data scientist career is not a big challenge for me. But I developed interest in science so i want to make a career in physics or geosciences. I cannot afford to start over from the beginning(bachelors), hence I am planning to do a masters in data science with physics or geoscience concentration. I am fine with combining machine learning and any physics stream as long as I do actual physics(or geosciences).
I would be interested to do phd but I need sometime in between to stabilize myself financially. That's the reason I asked whether any industry related jobs available which involves both machine learning and physics.I am totally out of guidance, and there is no clear path for me. I would be much grateful if someone would gives me a clear idea.

 

[quantknight]

would someone please advise on the query I raised?

 

[anorlunda]

Perhaps a university our a research center.

 

[mfb]

Many areas of physics have applications for machine learning, but it is unclear how many universities/research labs/companies will hire a machine learning expert without any background in physics for these tasks. Typically you have to understand the input and output in terms of physics. That is not too hard to learn, but it is additional training time. In addition, no physics background can make you less flexible in physics-focused jobs.

 

[Choppy]

Machine learning is a hot topic in medical physics right now. Consider imaging problems such as tissue classification and segmentation or tumour identification, computer-assisted diagnoses, problems in optimizing treatment workflow, radiotherapy treatment plan development, or treatment outcome prediction... these are all areas that stand to gain from machine learning.

And maybe this is pointing out the obvious, but, one of the things about most kinds of "research-support" related jobs is that they're not likely to be permanent. Someone may have a project in a field of physics that requires expertise in machine learning, and when they do and get some kind of grant funding, they'll hire on the expert, but eventually when the project is completed (or they run out of funding) they and you will have to move on. So aiming to support a particular field is probably not the best plan.

 

[quantknight]

Many areas of physics have applications for machine learning, but it is unclear how many universities/research labs/companies will hire a machine learning expert without any background in physics for these tasks. Typically you have to understand the input and output in terms of physics. That is not too hard to learn, but it is additional training time. In addition, no physics background can make you less flexible in physics-focused jobs.

Thanks for the response. I am expecting to learn physics along with my data science degree(2 years). Consider that if I pursue my machine learning degree along with physics, is that knowledge enough to pursue a career with physics? I am okay with extending my program's duration if I have to learn enough physics.

 

[quantknight]

Machine learning is a hot topic in medical physics right now. Consider imaging problems such as tissue classification and segmentation or tumour identification, computer-assisted diagnoses, problems in optimizing treatment workflow, radiotherapy treatment plan development, or treatment outcome prediction... these are all areas that stand to gain from machine learning.

And maybe this is pointing out the obvious, but, one of the things about most kinds of "research-support" related jobs is that they're not likely to be permanent. Someone may have a project in a field of physics that requires expertise in machine learning, and when they do and get some kind of grant funding, they'll hire on the expert, but eventually when the project is completed (or they run out of funding) they and you will have to move on. So aiming to support a particular field is probably not the best plan.

Thanks for the response, I haven't known about machine learning applications on medical physics until you said. I'll further research on the topic. Say if I study medical physics along with my degree for two years, would that be enough for me to get a job in a company or research lab? Is that enough knowledge to further pursue for my PhD?

On research part, yeah I read through our forums and Google and understand the reality of it , so as of now I haven't decided about that part. So I have fixed my goal as to complete the degree, and get a physics job(company or research lab).

 

[Choppy]

Say if I study medical physics along with my degree for two years, would that be enough for me to get a job in a company or research lab? Is that enough knowledge to further pursue for my PhD?

From what I understand so far, you've done a degree in something that's not physics. You're qualified to get into a master's program in data science or machine learning and want to enroll, but are interested in doing work in physics or geoscience areas, and want to know if you can pick up enough physics on the side to break into the field, and potentially enroll in a PhD.

Getting into a PhD in physics (or medical physics) generally requires a bachelor's degree in physics, or it's equivalent. This isn't the level of knowledge that can be typically picked up on the side, while studying something else. And in most MSc programs, you'll be pretty busy studying the MSc material. Most MSc programs don't allow you to pick up a minor anyway. Some medical physics programs are little more flexible in terms of the background of people they'll accept. So you could potentially look into doing a master's in medical physics that involves a machine learning project, but again, without a bachelor's degree in physics or something very close to it, this will be an uphill battle.

Getting hired to do machine learning work in one of these fields though, can depend a lot more on your experience in the area. If you were to get an MSc in machine learning and apply for a research support position on a project that's looking at say, using machine learning algorithm to reduce noise in CT image reconstruction, whether or not you get hired will probably depend a lot more on your ability to work with the specific algorithms and data sets that the research group needs you to work with than your general background medical physics.

 

[mfb]

Thanks for the response. I am expecting to learn physics along with my data science degree(2 years). Consider that if I pursue my machine learning degree along with physics, is that knowledge enough to pursue a career with physics? I am okay with extending my program's duration if I have to learn enough physics.

I don't know about medical physics, I only know how the jobs look like in particle physics and related fields.

Consider this list: Physics jobs and PhD positions somewhere close to particle physics that mention machine learning. Have a look at their requirements - you'll see a PhD (or MSc/BSc for PhD positions) in physics nearly everywhere. It is not impossible to apply with a different degree than listed there, but then you have to be an excellent candidate to have a chance. Typically these jobs have machine learning as a big part of the work, but not as only topic. In experimental physics it is expected that people contribute to the detector operation in some way, for example.

 

[quantknight]

From what I understand so far, you've done a degree in something that's not physics. You're qualified to get into a master's program in data science or machine learning and want to enroll, but are interested in doing work in physics or geoscience areas, and want to know if you can pick up enough physics on the side to break into the field, and potentially enroll in a PhD.

Getting into a PhD in physics (or medical physics) generally requires a bachelor's degree in physics, or it's equivalent. This isn't the level of knowledge that can be typically picked up on the side, while studying something else. And in most MSc programs, you'll be pretty busy studying the MSc material. Most MSc programs don't allow you to pick up a minor anyway. Some medical physics programs are little more flexible in terms of the background of people they'll accept. So you could potentially look into doing a master's in medical physics that involves a machine learning project, but again, without a bachelor's degree in physics or something very close to it, this will be an uphill battle.

Getting hired to do machine learning work in one of these fields though, can depend a lot more on your experience in the area. If you were to get an MSc in machine learning and apply for a research support position on a project that's looking at say, using machine learning algorithm to reduce noise in CT image reconstruction, whether or not you get hired will probably depend a lot more on your ability to work with the specific algorithms and data sets that the research group needs you to work with than your general background medical physics.

I understand, I came across few universities which have flexible curriculum in masters to have extra courses. Consider I got into university, and want to study machine learning with concentration of physics, how much (no of papers in physics) do I have to learn to get at least a job in physics industry? I gone through the link mfb provided, and there are requirements available for machine learning with phd in physics/maths and computer science as well.

 

[quantknight]

I don't know about medical physics, I only know how the jobs look like in particle physics and related fields.

Consider this list: Physics jobs and PhD positions somewhere close to particle physics that mention machine learning. Have a look at their requirements - you'll see a PhD (or MSc/BSc for PhD positions) in physics nearly everywhere. It is not impossible to apply with a different degree than listed there, but then you have to be an excellent candidate to have a chance. Typically these jobs have machine learning as a big part of the work, but not as only topic. In experimental physics it is expected that people contribute to the detector operation in some way, for example.

Thanks for this resource. This is good. I will look into it.

 

[Vanadium 50]

? If I take data science master's with minor in particle physics or astrophysics, what are the possibilities that I would be employed as a physics related data scientist?

Zero. Sorry, but that's how it is.

Mfb was too kind. Every single position (other than a graduate studentship) requires a PhD. Virtually all places that hire postdocs cannot hire someone who is not a postdoc into that position. This is because their grant application specifically mentions "postdoc", and you have to use the money the way you say you will. As far as industry, the amount of particle physics and astrophysics being done my industry is microscopic. As discussed above, expanding this into areas where there are commercial applications might change things, but otherwise, you are chasing jobs that do not exist.

 

[quantknight]

Zero. Sorry, but that's how it is.

Mfb was too kind. Every single position (other than a graduate studentship) requires a PhD. Virtually all places that hire postdocs cannot hire someone who is not a postdoc into that position. This is because their grant application specifically mentions "postdoc", and you have to use the money the way you say you will. As far as industry, the amount of particle physics and astrophysics being done my industry is microscopic. As discussed above, expanding this into areas where there are commercial applications might change things, but otherwise, you are chasing jobs that do not exist.

Thanks for the reply. What about other topics in physics? I guess there would be industry jobs in machine learning with geophysics, medical physics,quantum information or optical information,etc? I feel it would be much better to do data analysis of physics or geoscience, rather than banking products or customer behavior. I would be happy if there is a chance to proceed into research in physics, but if it is not possible, I am fine with physics industry jobs which requires masters level education. (I posted this question based on the fact that many universities offer curriculum data science with business, banking or AI specialization, so I wanted to know about having a specialization in physics or geoscience).

 

[Vanadium 50]

I really don't know where the jobs are - my point was that this is somewhere the jobs aren't.

That said, the basic strategy of "I don't want to do this and I don't want to do that" has a built in failure mode. We had someone some years back whose complaint was (exaggerating maybe a little) "I don't want to do energy, defense, finance, health or software, and I need to live in Nome, Alaska. This degree is worthless!" I'm a little worried that if you pitch yourself as someone who is an expert in domain science, most places will be able to find someone who is even more of an expert.

 

[ZapperZ]

Thanks for the reply. What about other topics in physics? I guess there would be industry jobs in machine learning with geophysics, medical physics,quantum information or optical information,etc? I feel it would be much better to do data analysis of physics or geoscience, rather than banking products or customer behavior. I would be happy if there is a chance to proceed into research in physics, but if it is not possible, I am fine with physics industry jobs which requires masters level education. (I posted this question based on the fact that many universities offer curriculum data science with business, banking or AI specialization, so I wanted to know about having a specialization in physics or geoscience).

Throughout this entire thread, I keep seeing this recurring theme that you think you have a solution waiting for a problem.

I know that what I've said is a useless comment for what you are seeking, but things normally work the other way around, and I don't recall ever reading about any funding or proposal where one is hunting for a problem to solve based on a solution at hand.

Zz.