What would you ask the MATLAB leadership team?
Hans Scharler
on 9 Oct 2024
Latest activity Reply by Mike Croucher
on 9 Feb 2025
Let's say you have a chance to ask the MATLAB leadership team any question. What would you ask them?
62 Comments
Time Descending
I am new to MatLab. Just started learning.
1- What are the best practice to learn it?
2 - How does the MATLAB leadership team plan to enhance community engagement and support for users, particularly in terms of educational resources and collaborative projects?
3 - In light of the rapid advancements in artificial intelligence and machine learning, how does the MATLAB leadership team intend to integrate these technologies into the platform to ensure it remains relevant and competitive in the evolving landscape?
Hello Drita,
1. Use it everyday! Find things to try like graphing data from a spreadsheet. Download datasets and work with them. 2. EdX and Coursera! They have official MathWorks course content. 3. Check the FileExchange out and Cody. You will find some community inputs that are valuable. 4. MathWorks is doing well to provide ML and DL capabilities. Some users have git repositories with LLM integration.
Cheers,
In my question I want to distinct from MATLAB teh product and m the language.
I was wondering, do you have a vision or any plan to overhaul the m language?
Since teh design of the language has a direct effect on the performance I wonder is the lnaguage and choices made 25-36 years ago make progress a bit more challenging now.
One way to have a "new" language but I guess you rightfully consider the current code of MATLAB to be a great asset.
So maybe some chnages (Opt In) + Decorators could be a good balance between progressing the language and performance and having backward compatibility?
Anyhow, it would be great to have some of your vision for the lmnaguage and the run time to be shared.
This reply was flagged by Stephen23
Assalomu alaykum Hurmatli Hans Sharler men Universitetda o`qituvchiman Allanazarov Doniyorman MATLAB dasturingizni tekinga variyantlari yo`qmi litsenziya chunki o`quvchilarimga o`rgatishim kerak. MATLAB 2024 versiyasida ishlamoqchi edik. Menga yordam bera olsamngiz sizdan minnnatdor bo`lardim.
Sorry, @Doniyor, there are no free versions of MATLAB. The Academic versions of MATLAB are relatively inexpensive (compared to the commercial versions.)
Would there ever be a consideration for having a Home Use version of certain product offerings that aren’t currently available? For example, Battery Toolbox? System Composer, Simulink Coder?
I ask because these tools are obviously cool but hidden behind a very expensive mountain. Case in point Stateflow only builds code if you have Coder. But if you simply an enthusiast that wants to learn and build something really cool with your Arduino, no joy.
I really like the Battery toolbox also because I am one those weird people that just like modeling batteries and understanding thermals for the pure fun of it. Having a limited output version of these tools would be a fair trade off. I would gladly pay a higher than $45 rate for “lite” versions of these tools but not thousands per year as I would never be able to afford that. Again I am an enthusiast. If I ever came up with something novel enough, I would gladly go into business and convert to the standard use license. It would very much worth it.
This is a difficult topic. There are multiple considerations.
One of the considerations is that Mathworks is concerned that if expensive products are made available to Home users at low costs, then commercial companies would claim to be Home users in order to get low-cost access to the products. This is apparently a fairly significant consideration.
A valid point.
By this logic I would be able to use my Home Use License commercially. I wouldn’t, to be clear. My recommendation was that perhaps MatLab and Simulink Coder functionality for Home Use Licenses, would be limited to Arduino for example. Not much can be taken seriously on the platform for that matter. Further, Items like System Composer and Battery Toolbox can be great tools if offered with little capability to output. But the most important points I would like to contend are:
1. Having long term access to these tools and learning how to use them while providing demonstrated capabilities as an individual can contribute to getting a job in a field that uses such tools and reduces the learning curve once hired. 2. Having access to such tools and knowing how they work and what wonderful things they can do translates into further licensing transitions once the user decides to go Commercial and the high costs are better justified.
I don’t want MathWorks to lose here. Rather, I want them to grow by means of not baring users from learning to leverage the tools by means of insurmountable costs. I am a self proclaimed MathWorks enthusiast and Evangelist. Therefore my desire is to see more use of the tools and to encourage our younger generations of engineers and scientists to implement them in their professional endeavors.
I am also a long-time MATLAB enthusiast and have been using MATLAB as a personal hobby tool for many years. Whether it's data analysis, algorithm development, or simple scientific computing, MATLAB has provided me with great convenience. However, I only use MATLAB for personal learning and non-commercial purposes, and I would never use it for any commercial activities. Therefore, the high annual licensing fee of thousands of dollars is naturally unacceptable to me. That said, if the price could be reduced to a range of a few dozen dollars, it would be very reasonable and completely acceptable. I hope MathWorks can take into account the needs of individual users and hobbyists like me and introduce more flexible and affordable pricing strategies. At the same time, as a MATLAB enthusiast and a supporter of the open-source community, I am more than willing to contribute to the promotion and advocacy of MATLAB. By attracting more individual users and open-source enthusiasts, the MATLAB user base will further expand, which will undoubtedly contribute to the healthy development and ecosystem prosperity of MATLAB.
Generally speaking, Mathworks offers a Home license, for use with personal projects. The base price for the Home license starts at about $US150, plus roughly $US45 per toolbox.
It is true, however, that a number of toolboxes are not available for Home license.
Depending on your geography, the Home license could cost as little as $95, plus $29 per toolbox:
There is also a [much cheaper] Student license that can be used if you are enrolled as a student and don't already have access to a campus-wide license that many schools have:
This is true. I’m just glad to get the licenses to the home use stuff. Better than Nada…
This is a great point.
Thanks for all of the questions. We answered some of these questions at the Community Advisory Board meeting, but I didn't get this thread updated. I am working on that now.
"Is it possible to work and code in Matlab for Matlab / the Mathworks team ?"
The standard schedule of releasing two versions of MATLAB per year is preceded by a pre-release—a very stable beta—wherein users can provide feedback on (hopefully minor) bugs before the public release.
On the other hand, the New Desktop has been in beta for multiple MATLAB releases, allowing users to provide feedback at an earlier stage that can influence more than just bug fixes, with the downside that some aspects of the beta have been in alpha or not yet implemented.
How do you think the New Desktop beta has gone? Is the average 3-star rating (on File Exchange) troubling or expected and why?
Do you expect MathWorks to provide additional opportunities in the future for users to provide feedback on early beta and/or alpha builds (on a more general scale than usability studies, as was done with the New Desktop)?
We have released several updates to the beta, each providing valuable feedback. Yes, I expect that we will share more betas and early access apps.
I have been using Matlab quite heavily in my work over the last 10 years or more. Area of work: quantitative models in economics (discrete dynamic programming problems, collocation method for solving functional equations, etc).
Matlab used to be a great complement to Fortran or C since it was slower but quite user-friendly. Now with Julia and Python+Numba: they are also user-friendly and moreover they are faster than Matlab (close to Fortran/C in terms of speed), and also cost-free. So in my personal view, Matlab should invest heavily in improving code performance. I nopticed that a great step in this direction was taken in 2015 with the new execution engine that extended a lot JIT compilation see e.g. here (loops-heavy codes that used to be really slow became faster). More needs to be done along this direction, in my view.
I think existing users might not switch from Matlab to Julia/Python+Numba since learning a new programming language requires paying some fixed cost (but Julia's syntax is very close to Matlab...). When it comes to new users, however, I see Matlab future kind of endangered. In many universities, computational course have changed from Matlab to Julia/Python and I see new cohorts of master and PhD students learning Julia/Python instead of Matlab.
Just to be clear, this is not to bash Matlab which remains a great language, but to point out some avenues for improvement from my area of expertise.
I totally agree with you.
Performance is the (Only?) feature everybody wants.
I hope for big news on improvements on the JIT engine.
I wonder what can be done on teh language syntax level to assist the improvement of performance. One easy feature I can think of, API for "By Refrence" in functions and add "Buffer" input for many functions to avoid allocations.
A lot of work is done on performance every release. Indeed, before I became the author of The MATLAB Blog, I pitched a 'MATLAB Performance' blog to management since I felt I had enough material to talk about performance alone. Since Loren was retiring I was offered The MATLAB blog where I can talk about performance as much as I like.
I am obsessed with performance in MATLAB and do a lot of work with development in finding where we should be faster and communicating where we have got faster. Here's an example story From HPC consultancy to a faster fzero function in MATLAB R2023a » The MATLAB Blog - MATLAB & Simulink
Before I joined MathWorks, I was a Research Software Engineer in many languages including Python (I even taught High Performance Python!), R, Julia, C++ and, of course, MATLAB.
I recently worked with a researcher in Maths where we used Python and MATLAB combined to help improve an algorithm way more than would have been possible if we had chosen either language alone. Check out the story on MATLAB Expo https://www.mathworks.com/videos/the-classix-story-developing-the-same-algorithm-in-matlab-and-python-simultaneously-1730890135715.html
Anyway....I'd love to hear from you what you think is slow in MATLAB. Message me here, or on BlueSky Mike Croucher (@walkingrandomly.bsky.social) — Bluesky and we can swap email addresses and take it from there if you like
Arranging to update or reinstall Matlab through my employing institution continues to be very difficult, and requires the direct participation of our system admins every time it needs to be done. They insist that they must perform the installation directly, to ensure every individual or lab is up-to-date on their license payments. Is there any way to make this easier, for the admins and for the user?
For example, Microsoft seems to be able to make its products downladable directly to the user through Single Sign-On authentication, without the need for their institution admins to be involved.
Mathworks is willing to align payment schedules over a number of licenses. That would reduce the burden of checking license expiry dates to once (initial installation).
There isn't any need to continually check license payments anyhow, since an unpaid installation will simply expire.
... Unless, that is, you are running with centralized license servers and all of the payments being discussed are internal...
There are several ways that corporate customers can deploy Matlab. What you describe is unique to your company, and the solution you describe is similar to how it works at mine. It's the Admins who decide - although it's worth noting that different configurations come with different pricetags and limitations. (For what it's worth, it also sounds like your employer has a base restriction on downloads or installs. Anyone can download MATLAB from the website for free. It simply won't run without a valid license file.)
The question I'd ask them "over a beer" is: With competition from the more rapidly growing (and supposedly more efficient) languages like python and Julia, what are your long-term plans to not go extinct the same way as IDL did?
(Just to make it very clear: This is not intended to be an agressive, malicious, meanspirited or evil question. Since students at our university and as far as I understand most other universities now have python as their main language for "programming in sciences", and as they move into the workforces so will the use of python where previously matlab was used. Therefore this is the question I'm most curious and worried about.)
Great question - definitely hard to answer succinctly without starting all sorts of holy wars, so I'll try to take a high level stab:
- Make sure the experience of using MATLAB continues to be great. As owners of the full stack (IDE, language, execution engine, online platform, core libraries, ...) we are uniquely positioned to tie everything together into a coherent system. On a personal note, I just transferred from Head of Product for MATLAB to "head of MATLABiness" to focus on precisely this.
- Make sure MATLAB is great for what MATLAB should be great for - engineers and scientists doing engineering and science. MATLAB is a very general-purpose environment for engineers and scientists, but a very specialized environment for software developers and IT folks. With it's best friend Simulink, MATLAB particularly excels at supporting the end-to-end workflows of developing engineered systems.
- Ensure users can use the right tool for the right job. For the last many years, this has often translated into driving a great experience of using the best of MATLAB with the best of Python together.
- Keep making it easier and easier to access MATLAB. This means technology (MATLAB Online, making it easier to work in cloud environments, ...) and licensing (Campus Wide License, ...)
We of course continue to keep a close on the great things happening with other languages to figure out if there are important gaps in MATLAB or opportunities to improve integration.
That's as fair of an answer I'd hope for in public. (But now I get even more curious about what less official discussion I might hear over that hypothetical and figurative beer...)
Why not grabing an actual beer with Michelle? :)
For the near future geography makes that challenging - an ocean and a couple of time-zones and such...
Yet I feel it mostly say you're focusing on MATLAB's strong points.
We also wonder if you have vision for the weakpoints.
Specifically:
- Performance.
- Low Level Memory Management (Avoid allocations).
The vertical integration is indeed the strongest point of MATLAB and probably will make it first choice for many (I included). Yet we'd be happy to see improvements on the weak points.
Is Python still considered open-source, now that Anaconda charges licensing fees?
Isn't that more or less the same scheme (is that how one spells the word pronounced as skam?) as "red-had enterprice" with an increasingly restricted pay-for-service of something that are more or less a packaging of software available free from somewhere else, or have I missed something? If I was a python-user I would start to worry about that.
That's a great question. I think Matlab should keep on improving the performance thanks to JIT compiler. The two closest competitors (Julia and Python with numba) are already faster than Matlab, besides being open source. In my computational work, speed is important and I use Matlab/Fortran. I am considering switching to Julia. However I noticed that after R2015 Matlab underwent a significant improvement.
In my vision some decorators will be added ot the core language to assist achieving Numba performance in MATLAB.
Is there any reason why release of certain R2024b site license types would have been delayed? My institution's (Dana Farber Cancer Institute/Mass General Brigham) license admins claim that R2024b is not yet available for installation on our computers. They are only able to distribute R2024a.
If it's network licenses... The 2024b network license has an update to the FlexLM that breaks some server setups. This is made by Reneva, not Mathworks, and the only want to get 2024b+ to work if affected is to change the license setup.
("Reneva" should be "Revenera")
I recommend that you have your license admins contact your sales representative and/or Customer Service and work with them to determine why R2024b is not yet available for installation on your machines. I am not aware of any technical limitations that would prevent this, but then again I don't do very much with licensing (mainly just clicking Accept when I install MATLAB myself.)
That may be due to an internal corporate policy. I have had to use a prior release (at multiple companies, sometimes by up to eight releases) due to the company not having yet paid to upgrade or some IT policy about approving applications or upgrades.
In my experience, official training materials from MathWorks, such as Onramps and all-day courses (virtual or in-person), are high quality and very helpful. However, there are only so many Onramps available while courses can be expensive. What ideas do you have for making MATLAB training materials more accessible to the average user?
There are several things we do here:
- We continue to invest very heavily in developing and maintaining free onramps. There are currently 24 of them.
- We partner with online learning providers like EdX and Coursera to support the development of massive open online courses. We support professors who develop courses, and we even lead the development of some.
- We offer the Online Training Suite, which gives access to all of our online course content. I've lost track of the details, but I'm pretty sure this is included with Campus Wide License and can be bought for an entire organization.
- We host open courseware for over 500 courses, about half developed by MathWorks and half by other users.
Do you think that the current number of MATLAB toolboxes is about right? If not, should there be more (e.g., either due to adding new toolboxes or splitting existing toolboxes) or less (e.g., due to combining existing toolboxes)? Whatever your answer, why do you think that?
In particular, I find the partitioning of certain deep learning utiltiies across different toolboxes to be a bit frustrating. For example, that unet is only available with the Computer Vision Toolbox. I think a Deep Learning Toolbox license ought to be sufficient to access unet(), or any standard network architecture.
Can you comment on how these partitioning decisions are made, when you have closely related toolboxes like Computer Vision, Image Processing, and Deep Learning?
Personally I think there are a lot too many toolboxes... close to 75 of them at present.
Starting with Matt's question - there are ongoing conversations about where the boundaries are and should be between any given set of closely related products. Sometimes we move functionality into a more broadly used product, sometimes we aggregate a bunch of specialized functionality into an add-on product.
We moved a bunch of very common stats and signal capabilities to MATLAB a handful of years ago (think things like the "omitnan" flag and islocalmax), and continue to keep an eye on other opportunities.
Ideally, the product split represents the right alignment of a logical grouping of capabilities that provide good value to users. It's not always perfect - sometimes a team builds something because they need it ASAP, and then we need to figure out how to make it more widely available. Sometimes we can move functionality, but sometimes it might be so reliant on other code in a toolbox to make it difficult to move.
To Grant's question, I think we've got a reasonable set of core toolboxes like Signal Processing Toolbox and Statistics and Machine Learning Toolbox. There's a lot of nuance as you get more niche that make it trickier to get just right (as Matt's comments point to).
Have you considered making all the infrstructure toolboxes free (Beside Coder) and then toolboxes should only be seperated by the problems they solve?
For instance, GPU, Parallel, etc... seems like basic features in a modern language these days.
I noticed yesterday that Mapping Toolbox is now listed under RADAR umberella. That feels quite wrong to me.
If you're talking about its placement on the Products page, it is indeed in the Radar section under Application Products. Looking at the other application products, I'm not sure if there's a better place for it among the existing sections. It also doesn't really seem to fit in any of the groups under either MATLAB Product Family or Simulink Product Family.
If you could move it to somewhere else, where would you place it and why?
Keep the questions coming! I'm happy to answer as many as I can.
Michelle has entered the chat... :)
Is this topic open for everyone to see, or limited to certain users?
If it's open, we should refrain from mentioning anything that's not public knowledge
Hi Yair. This is an open discussion.
MATLAB is continually developed for Windows, Mac, and Linux. Then, MATLAB Online and the New Desktop (currently in beta) were added. Will development for these new platforms lead to a consistently higher development workload for the semiannual releases, or will it lead (eventually, if not already) to more efficient cross-platform development and consistency between platforms?
We've got a great setup that means most developers, most of the time, don't need to think about platforms. Our continuous integration and test system simultaneously builds for our platforms, so new code that is submitted is verified to work across the board.
The biggest hot spots that might make it harder for a developer to support multiple platforms are third-party libraries and UIs. Adding support for a new platform (like when we added native support for Apple Silicon) requires getting reliable, performant builds of all of the necessary third-party libraries on that platform. That's a bunch of work, but once it is set up the maintenance is often quite manageable. UIs were tricky for years, because we used Java on the desktop operating systems and JavaScript on MATLAB Online. We've moved all of our UI development to JavaScript now, which gives us a single UI code base that works on all platforms and MATLAB Online (MATLAB Mobile is still quite custom, but the work is localized to a focused team). This will allow developers to again largely not have to worry about platform. It's not perfect, because each OS has it's quirks that we need to accommodate, but it's manageable.
This is a helpful explanation. I think that the transition from Java to Javascript will help with long-term sustainability and macOS compatability.
If you could immediately add to or change anything in MATLAB, what would that be and why?
Oof, that's a tough one. I live in the world of making things better, which means I spend a lot of time thinking about things that aren't serving our users as well as I think they can be. I'm going to dodge this question ... maybe over a beer at the next in-person community advisory board (Sorry I had to leave early last time!)
What fraction of new features and functionality are derived directly from user feedback?
Ultimately, everything we work is because we think it's something that will help users. An awful lot of this is because we've heard it from you (the royal "you", not just Grant!). This can take a lot of different forms, and depends in large part on the nature of the product. The development of small, new products can be very heavily driven by the specific needs of a small set of users. We can iterate pretty closely with some of these users to ensure we are tackling the right problems in the right way.
As you can imagine, MATLAB is quite different. There's no small set of users who can guide the whole ship. When I was shepherding MATLAB (2007-2024), I tried to help us develop a strategy that lined up broad market trends with specific needs of our users. Big ticket items should help make MATLAB more successful and address real challenges we've heard from real users. But the bigger the item, the less likely it looks just like something any given user asked for.
That said, most features in most MATLAB releases are smaller and often more incremental. These are much more likely to directly reflect what we are hearing from users. We synthesize a ton of sources as we plan each release to try to get a sense of the best impact things to work on. Source include: direct feedback from users (from the community or via our field team), MATLAB Answers traffic and activity, tech support activity, input from different user advisory boards (we have an increasing number of these focused on different parts of our products), input from key users of functionality, usage data, etc ...
How long does it usually take for a new feature to be released, counting time from an idea being suggested (by MathWorkers or users) to being included in the public release?
To answer this, it helps to understand how we plan:
We just went through planning for R2025b in early November. Just about every feature coming in R2025b should have been accounted for in the plan. With our usual release schedule, that means the idea had to make it to a plan that was developed about 10 months before the release. Of course, lots of features take longer.
Usually only small quite small things (often bug fixes or very minor enhancements that feel like bug fixes) can be introduced and addressed in less time.
Sign in to participate
