Bestand wählen
Merken

Key Note Lecture: Managing research software development - better software, better research

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
thank you very much to the organizers for inviting me along today to talk to you all and I'm hoping that's
not just I'm not just to be talking to you you're going to be giving me information so that I can use this in future talks and steal although the credit for it in the director of something in the UK called the software sustainability institutes and my role and my institutes role is really to help researchers understand how to improve the way that they are using software and that might be in a using software for simulation purposes for data analysis
it is about to developing software it is about all of the different ways in which
software is a part of the life of a researcher today and it's been really
good this morning seeing lots of different examples of the ways in which people are using software for a modeling and complex interfaces for modeling and differing types of fluid dynamics so I you can tell I was in this room and and it and seeing all of the different kinds of technologies as well that they're using so I have a car or a quiz for you a question who can tell me what is the most popular modeling and simulation tool used in the world today so who thinks it's sad evinces Matlab OK
and who thinks that people do this using something like C plus plus no no and no 1 thinks the suppose plus python you see you know what I'm talking about the XL so
they are most popular and modeling and simulation package in the world is itself because is used widely in geez
wide in finance its use widely in a large range of different research subjects and
it's often use because it is a combination of a database a simulation package and analysis package and a visualization package although really it's none of these and but can anyone here has anyone seen the spreadsheet before and this is a very famous spreadsheet this is the Rinehart and Rogoff spreadsheet it is a set of data and analysis that accompanies a paper by 2 economists Rinehart and Rogoff which has been used by a lot of different governments to understand that the way that different economic measures and change the way that's basically had different economic principles would change the way that you can measure different countries' economies so can you spot the mistake anyone see it I give you a clue stand here so you see here there's an average here that says average L 32 L 44 and what we notice here this is l 30 he L 44 what we notice is missing the countries missing here so what you have here is a classic example of the types of mistakes that you can get particularly when you use excel as your modeling platform and none of you use Excel as a modeling platform so it's OK and that the prom with this is that it's an example of how EU errors can have really large impacts on places the you don't really expect them so in this case and because of this mistake what you ended up with is a whole set of slightly different calculations on something that is being used by a lot of different governments to set their policies so in Germany and the UK and so on all of these are governments have used this spreadsheet the results of the spreadsheet to decide what rate of what rate of taxation used you might have and so on and actually when you look at it the difference you get from not putting in these columns is fairly small because it you know a lot of these don't really matter and Sony Belgian that really really matters in this calculation but the problem here is it's 1 of perception it's 1 of how people can see the truth in this and this is basically down to this problem of reproducibility right so why do we want better research sorry better software we want it because we want better research and the reason is so that we can trust the results we get and this example was it is in reality something that none of you will be subjected to because you don't use excel and is an example of how trust in science is eroded because people can't understand where the results are coming from and so you might think that it's only people who use excel that have this problem it's
not so you can see other examples here so he is something where and this is Jeffrey Chang working in effectively molecular structure has done a lot of great work in terms of understanding you know he's 1 he was 1 of the pioneers of a new movement of understanding molecular structure and coming up with a different ways of identifying it and the problem here was that he had been using I love the way that this this basically this news reports as a home a data analysis program by homemade data analysis program he just means a piece of software that his research group had written so that's the same as all of the things you are writing be careful this is how science journalist will have a report your work and which it flipped 2 columns so very very similar problem to the 1 that we saw in the 1st example where instead of not something all of 1 column what's happened is 2 columns have simply been interchange in the data file and and as a result he had to retract a whole the papers and the problem here is that that's a very small error for what is otherwise a very good body of work so the date is all good the bit that he made the great strides in the place where he made the most advances is in the gathering of the data in getting high-quality data but because his software was seen to be wrong that means that his data is also potentially seem to be wrong so when you're govern good quality data or even if you're working with good quality simulations elsewhere 1 mistake in your pipeline can cause your problem
and and another example here and the people a PhD students in this audience OK this this 1 is for you and this is a paper from Lorente itself that was published in science and it is about the migration of peoples from Africa into Europe and also what their discovery made was that it appeared that there was also a a migration of people from Europe back into Africa and this is an example where again you've got a large consortium you've got some great new data so they developed a new technique for extracting genetic material from very old skeletons bone samples the gave them a new dataset that they were able to run through a new fruit through an existing pipeline to analyze the way in which genetic material propagated between different basically peoples as they move around and they got this wonderful new result the rent is the PhD student who is the 1st named offer on this work and but it was wrong so that the issue is that because it was new data they hand check to see what would happen when they put it for the pipeline and 1 stage in the pipeline accidentally threw away information because it looks like it wasn't the right information so again what you ended up with was some very good work of a very large consortium written with a leadoff who was a PhD students and being suspect because of 1 error in the software and this is a good example of how you fix it though because they made all of their software data available and so people found out that's also the kind of the thing to to understand from the very 1st spread she error message I gave people found out about this because they made their data available and was happy to answer questions on it the problem is that in many subjects that's not the case and
so this diagram looks very complicated but it is an example of a wide scale survey done by some people in computer science all of papers in the Association for Computing Machinery's Computer Science journals and proceedings of conferences and what this shows basically is the different ways in which can fail to get hold of the code for a piece of paper that's been published in a computer science journal so the 601 papers that went in and you get a whole load that fall out because you can't find them you can't get in touch with the author you can't the off will not provide you with the that piece of software the build fails and so on so in the end you only you only end up with a very small number of that actually show you where the software is 85 you I did find it through other places but even of those you still get a few where you can't even though the software so even in a subject like computer science where you would hope that the practices were very good a ready the issue is that quite often you just do not see the code and therefore you can't identify whether or not the results are good in the 1st place so that's the scare story that's that's the reason why I'm here talking to you so the question is is this
true across all of software and the answer is yes so at the institute I worked for 1 of the things we have been trying to do is understand where Is software used and the answer is is used everywhere so we surveyed researchers in the UK to identify how much software is being used and the answer is basically it's being used almost everywhere so the key finding here is that 68 per cent of people said that they would not be able to do their research without software in there and so it's not just CASE making a harder than it is impossible for them to do their research and and a lot of them develop their own software so over half of them are extending or developing new software but only only a few of them so less than 30 % have any had any formal training in doing this so we're not wearing a s a country here where everyone is using software to do their research and no 1 has been trained on how to use it or how to develop and the same is true in other countries so a similar survey was done in the United States giving similar numbers and they're now surveys that looking at what's happening in places like the Netherlands and also this work being done to understand more of it in the in Germany which I'm hoping staff will be able to talk a little bit on the panel later about so this is basically saying the software is everywhere you can see elsewhere so if you look at
Nature papers and look to see mentions of software they're very hard to find but if you simply read through them look you can see you can see tools being used everywhere so on average a nature paper has 7 mentions of different pieces of software in each 1 so what this is saying is that scientists well aware that they're using software and they're well aware that and it's dependent on it and the question then is you know why do they find it so
difficult so that that is a good question and in general as researchers find it difficult to use software and it's a little at all because I'm sitting in a room here where a lot of you are experts in using software so the question would be why do you care why why is it the case that you should worry about this about the fact that most researchers find it very difficult to software what a different reason so 1st of all you can look at the root of the reasons why people find it difficult to and so here's an
example of some work that was done by Victoria Stodden to identify the barriers and a lot of it is a runtime so it's time to documenting cleanup software dealing with questions from users and so what else competitors gaining advantages all of these things are the sorts of fears that come from where you're in an environment where competition is really difficult so publish-or-perish is something that a lot of people say and when you look at the different challenges here they're all related to publish or perish so time to documenting cleanup is really saying I should be writing another paper and dealing with questions from users really saying I should be writing another paper and competitors get Mingasson advantages really saying I should write another paper but the issue here is that all of these things could be seen as as not challenges not barriers but enablers so dealing with questions from users that is the same as saying I'm getting new collaborators with my research documenting cleaning up means that other people are using myself software and perhaps I can get more citations and more knowledge some papers competitors may get an advantage sure that is 1 of the ones which particularly early career scientists early career researchers and say is 1 of the biggest problems but actually and it's something that doesn't happen on awful lot particularly if you turn those competitors and collaborators so the question here is how do we overcome those challenges and persuade more people to invest time and effort into the software that they're producing
and here is perhaps the reason why it becomes hard for people to do this and the reason why I am talking to you to try and make a little bit of a change and that is that in research it's all about your reputation and and 1 of the biggest problems that we have is that we live in a community where it's very easy to basically and be critical we're trained to do it we're trained to be critical when we do reviews of papers which aimed to be critical when we come and see a seminar and we are trained to be critical when it comes to publishing of any of our data on models or or software and that's fine but it can go to far so in this particular case the problem here is that people's levels of criticism on not necessarily well balanced with the thing they're criticizing so someone is publishing a piece of code so that they can make it more available to people perhaps you shouldn't basically be swearing at them and basically telling the person that they're coding is is dreadful it is terrible and that they will never work in this industry again and so on given the reputation is everything as a researcher how do we encourage people to do something right rather than shout at them when they do something wrong so this is what I would
challenge you all to do I would challenge you to understand the things that you can do and you can take back to your groups and persuade them to do to make it easier for the people who come on after you so as to make it easier for all of your PhD students and master students to make it easier for the people that you are going to employ in the future for the people you're going to be doing research with and it's only 5 things so the first one is train yourself in your team people of heard
all software carpentry or data carpentry excellent a few of you so software cotton in data carpentry are effectively an open source movement for training researchers they provide 3 courses for the basics that you need to do private data analysis and office software development so their teaching people to be confident enough to ask the right questions that's all it is and you are asking people to to learn not to the experts that but to be good enough that they don't feel scared about asking questions and what's
your buying into his basically a global community of practice so software carpentry and data cottage you've never been taught on all 7 continents and earlier this year we had our 1st workshop in Antarctica so we we finally have the full set but the more importantly there almost certainly a workshop going on near you if you in Europe and it is a place where there workshops running almost every week and more and more institutions are starting to use this as a way of teaching the basic techniques in being a good computational research on the just being a good researcher to all of the people at the institution so if your institution is not providing and these courses ask them why if it's because they're already providing other courses that do the same job that's great if it's because they didn't think was important tell them why it's important because if you don't teach people about these new skills particularly when their students it's light not teaching students about basic statistical tests it's like not teaching students how to give a good presentation and it's 1 of those basic skills you know cannot do research without so make sure that all of the students and younger researchers early career researchers at your organizations are getting and so that's the first one the 2nd 1 right to
strangers and so what do I mean by right for strangers when we look at reproducibility how 1 of the largest set of work has been done in the mathematics community in the statistics community and 1 of the people
who's been doing that has been a person called David Donahoe and he has been looking at this in trying to understand the way that we can produce code and why it's important to make it easier to understand and 1 of the things that he he is coined he's mentioned is that although it might seem unnatural to help strangers and really what we mean by stranger is anyone who doesn't possess OCR short term memory and experiences so what does that mean
who who doesn't and possess your you know short term memory experiences just people reading your paper so people reading your paper don't understand what you were thinking and future research collaborators so people you've not been working with don't understand your thinking and the people who are reviewing your paper or your grant proposal they don't know what you're thinking and future members of your research group your current students recall offers they don't necessarily know all of what was in your head as you were writing that code and importantly you as an author I don't know what you're thinking about 6 months afterward so when we talk about writing to strangers we don't actually mean just anyone what we're saying is a stranger is all of the people who might actually make use of your research and your research software in the future and the most popular person is you so what can you do and
so I was a co author on a paper called best practices for scientific computing and with a lot of other people which goes into the the evidence behind why different approaches are good or bad for software development and how they can be applied to scientific computing so if you are really 1 to look and see why for instance we don't necessarily recommend test-driven development but we do recommend testing this is the sort of paper for you but the problem is that when we produce this paper almost no earlier and no 1 was able to put this into practice so to do all of the things in this box here I was impossible to almost all research groups so a subset of people from here
when produce this paper which is good enough practices in scientific computing so this is the set of lists that you can go and you can talk to your new students and say don't worry
about all of these are just worry about these things here and what I what I have in here and they're really really
obvious ones but the ones that that help in the long term and help you make beer more coffin researcher so organ say very much about data but quite often I like doing this as a set of examples of how good people's practices are particularly in a crowd which is more for software so how many of you back up your role or data I guess less so because you're creating the data DU DU kind of firms submit your data to repository yes OK and get to do your wife and the people know idealized X and that's good so there are a whole lot of things around data the ones we're interested in and software and so
what has been what has been articulated in this paper on good enough practices is some basic guidance on how to develop software in a better way and it's interesting because 1 of the things that it doesn't do is sort of say you must do it in a particular way so for instance use version control the good enough practices paper says you must use version control it accepts that some people start doing version control by naming their files with different names and that is version control most of us would laugh at that and go you know that's that's not good enough but it represents a significant step forward from calling all of your programs a daughter and so what we're trying to do here is improve practice gradually and documenting for future self when those important ones is learning to be modular so 1 of the things that we are trying to teach people is splitting things up into a small pieces because they are easier to understand and also to give to other people and 1 of the problems with this is that particularly when we're working with code that has been produced by other people it can be very large it can be tens of thousands or hundreds of thousands of lines it can be hard to understand so I think the key thing that we we try and teach people is learning to be modular learning that things up and 1 of the key challenges for you is understanding how you can teach that principle to the people you work with your new students your new collaborators and given that the thing that they're learning to make modular is your code so think about your own code and think about how you could make it more modular and the other thing is not doing it alone so 1 of the key things here is that we try and tell people to work with others because sometimes it's very hard to see what is going wrong without asking someone else to try and explain it back to you and that's coming from the basic principle of software engineering of code reviews you're looking for someone else to try and explain your code to you so you understand where it is you've not managed to explain your code to them and there are there's an
interesting initiatives that take this further so what I had on the last slide is the very basics what you want everybody to be able to do at some level but a lot of communities are doing this much better at taking a further and we're here a little bit I think about that in 1 of the torques that's following and but I wanted to kind of highlight the way in which they're doing this so a very good example here is from the earth sciences the set if you go along to this this link here they will show you the guidelines that they have developed over a period of about a year and a half by talking to all of the people in the community understanding what are the challenges were understanding what they were a ready doing understanding what they wanted to do better and what they felt they didn't have the time to do better and they came up with a set of guidelines which has been accepted and approved by all of the people developing code in that community so rather than going from the top and saying you must do these best practices they went from the bottom and said what are you willing to make a change on and what do you think we need to make a change on and the result is the set of guidelines here and other people being doing the same thing so Lixia have been doing the same thing looking at how they can understand what practices that will help improve they're working by informatics Claria for the arts and humanities in image processing so the question I have for you here is what's the right way of doing this for this community and is there a role for the Leibnitz network to help do this and again I think we're a little bit more about this later was writing for
strangers that 1st step is developing a software management plan so how many people have heard of a data management plan OK so if you you give you have not so data management plan is basically something that identifies the types of data that you might produce as part of a research project was part of any project identifies who's responsible for them and basically says how you're going to publish them and keep them up-to-date and a software management
plan is essentially the same thing and the reason why software management plans a useful is because most recent software development is not planned so we don't we don't plan it happens because we're trying to do something I'm trying to trying to get at a new research results and trying to understand the new phenomena or something like that so it's evolved Ravan planned and so soft imagine plans a just a way of thinking through the process of running a research softer development project if you
won't find a more I've got a link for guidance and it's basically set a questions you can use to talk through the process of how U.S. develop your software but this really is what it looks like so this
is the smallest plan that you can really right that's useful so the question is can you say what your software will do and and for yourself or have a name and if you can't do that then perhaps should take step back but those I think anyone will be able to do and then use I get into the the more interesting ones so could intended uses of your software or is it just for you is it for you and your colleagues due have collaborators in other universities and what's the expertise that's required from each of the sets of users is this something which will only be used by people who are experts in the scientific field or is it more general is it something where you expect them to have experience the particular programming language or library and how we make the software available so I you publishing this on a website using a repository you just e-mailing it to people and this is the really interesting 1 how we'll software contribute to research we should all be able to do and how we you measure its contribution so how do you understand where the software using is having an impact and a very obvious way you mites do this in a plan is simply to say I will measure its contribution by the number of papers the by producing a year that use the software or maybe the number of papers of produced in the year have you use the software but as you start going along that line of questions and thinking about how to plan this you start asking how will I know how people use my software to produce a paper and that's getting into the topic of software citation and which I won't say very much about but again there is a talk coming later on the looks more into this and workflows around that so force that
is publishing your code so you can't do very much if the code is isn't out there and
so we talked about reproducibility and the key thing with firm reproducibility is allowing people to inspect your code and improves improving transparency improves understanding most importantly it improves trust and from the examples I gave at the start of this talk the biggest thing is around understanding whether or not someone trusts your results and 1 of the easiest ways of of giving them that trust is to make your code available and and that's becoming very
easy so there are many more ways you can publish your code if you're just talking about the code developments there are many different infrastructures that a freely available in your university may run a GitLab server for instance you may be using get hardball bit buckets to store your code likewise there are many different repositories that you can use to store archive versions of the code possibly are the most relevant here is a nodal run by certain which allows deposits of research software and allows import from I get how project straight into as a nodal collection so it's very easy to do and the other thing is that as part of this general workflow you're not getting lots more tools that weren't available and previously to help you test your code and make it available on different systems and platforms it's become much easier to publish your code for others to use because there's been a need from industry to do the same thing as open source software has risen in popularity so the tools and infrastructure around it have as well and the good thing is that as researchers and people developing research software it's very easy for us to benefit from this so make sure you publish your code and and and make sure you get and
give credit for software so they're different ways you can do this and the way that is closest to the way that we currently do things for pieces of research is by publishing a paper so there are no a large set of journals that will allow you to publish papers about your software as so if you want to do that you can publish a paper and then you're able to cite it in your research papers as well and there are also efforts to try and make this a direct citation so rather than going through this paper which as a proxy for the piece of software you can cite the software directly and the software citation implementation Working Group have published a set of guidelines and are now working with publishers to see how we can do this embedded into the scholarly communication systems and some of you will have seen I had an orchid on the 1st slide this is a researcher idea and I encourage you all to be using them and it simply makes it easier for machines to understand what you have published and how to associate everything so it means it you can benefit from tools make it easier to understand the impact of your work and some of these are things like all metrics and so you can we tools such as impact story to identify where your software is being used and what it has itself achieved and but I think what I really want to concentrate on is this is the biggest place where you can make a change in the way that practice is being done today and that's by being a better reviewer so how many of you review for journals or conferences OK so at least half of you so for all of you and then everyone else will do this at some point and there are 2 or 3 things you can do and that will help improve the whole of research culture as a whole the first one is they're all has not shown you where you can get hold all of their models their software all their data ask them the question about why they haven't and we often recommend you to send a polite and review that says I'm rejecting this but I expect this paper to be acceptable after major revision after revisions and but we would like to know where the software and data is and so that might be because by providing it to the journal to loud reviews to inspect if it's and software that is not open-source what might be publishing it on a repository if it is an open source and giving guidance about why this is important so the reviewer should really and the I think that the as a reviewer your aim is to give such good feedback and such good advice to the author of the they will want to invite you want to be a call for onto the paper and I've seen this happen once in 1 of the journals I edit where eventually the reviewer had given so much advice that they became a co-author and we had to get a new review have but a the constructive in your criticism so as as a reviewer you're trying to support people to become better researches and 1 of the ways of doing that is asking them to see the codon data the other is being constructed in your critters of says and the 3rd thing the 3rd and final thing is if you are a senior member and on on an editorial board or the editor of a journal make sure that all your reviewers know about this and tell them to do this as well because that's the way that we change things and is by effectively asking editorial boards to change as they are probably the key the gatekeepers of how we do research today so the last day on this publishing code is if you are interested in going further there's something called literate programming which take the approach the traditional papers adjust advertisements for the research you're doing and instead they enable others to look and inspect the code and do things with it so to 2 that effectively play with it
and a really good example of this is the like examples so for gravitational waves they produce a paper and but they also produced a Jupiter notebook the has all of the analysis code that will allow you to recreate and many of the graphs and results in this paper so lets you identify the high points and see how they did it and if nothing else it's a really good way of showing the others how you've done your work and getting an interested in your work so I highly recommend in that and last 1 get
expert help so what I mean by this as researchers we often don't have enough time to do everything we would like to do because we are trying to publish papers and the good thing that we have seen more recently is an acceptance by funders that it is a legitimate thing to have specialists in research groups or a contract doing to basically do the soft the hard software bits all the boring software bits for your project and what I mean by
this is is a term that in the UK is called research software engineers and which are a group of people specializing in and the development of software with in research and here we're also seeing this coming up in Germany so I German chapter has been launched and they are going full steam ahead and staff and is the person you want to talk to you if you want to get more information about that the key thing here is trying to get your organizations to recognize this role and to campaign to get more people to the to produce a career path for people to succeed in this role because I think that's the only way in which we will do things better is by ensuring we have all of the right people to do things better and so again it's very rare now in in biology lab for you not to have someone who's good at statistics and or someone who is excellent lab technician for someone who can basically do all of the little things with spinning bottles that I do not understand in the same way if you're doing a computational modeling projects who were the specialists you need to get into your team to really expand what you can do so that you can you can go and work on larger projects a greater scales and research software engineers a 1 of these roles
so I think that the main thing that I I'm going to kind of give you is a message from this talk is the problems we saw this in in research around the transparency the reproducibility and the trust in software all because people think of software is this black box that is rarely seen and is hard to understand so the key thing here is not to be afraid of letting other people see your software whether that's by publishing it whether that's by making it easier to understand whether it is just showing it to someone else and that is the thing that I would like you to take to the people you talk to me and say it's OK and no 1 will criticize you and if they do criticize you I will go back to them and say that they were wrong to criticize you and because if you don't do that people don't share their software have people don't share the software we will not get to a position where we have the reproducible research research that is of high quality and most importantly research that other people can build on the news so those are my 5
steps that is a 6 1 and and that 6 1 is keep on improving so there are many many other ways in which you can you can improve and I'm not going to go into them they're they're in my slides if you won't find out the things around legacy code
that we we see from industry and
the open-source community has had a lot of valuable lessons about project management and encouraging communities and there's a
whole body of work in understanding how you measure projects and understand their success or failure and and the
whole set of resources out there that you can use and because you're not alone and loads of people have been doing work in this area and the key thing is identified how you can transfer that knowledge on to other people so he would find out more about
my institutes and you can look at this all of these different links and but
apart from that I will leave you with the 6 steps to better research and I hope that you take them out back to your own organizations and tell everyone about them thank you very much
Datenanalyse
Dynamisches System
Subtraktion
Numerisches Modell
Fluid
Mereologie
Rechenschieber
Fläche
Komplex <Algebra>
Numerisches Modell
Subtraktion
Spannweite <Stochastik>
Numerisches Modell
Innerer Punkt
Leck
Resultante
Subtraktion
Sterbeziffer
Konvexe Hülle
Finite-Elemente-Methode
Datenanalyse
Taupunkt
Güte der Anpassung
Gruppenkeim
Kombinator
Extrempunkt
Rechnen
Term
Analysis
Stichprobenfehler
Numerisches Modell
Menge
Rechter Winkel
Mittelwert
Explosion <Stochastik>
Einflussgröße
Numerisches Modell
Analysis
Resultante
Zentrische Streckung
Explosion <Stochastik>
Stichprobenfehler
t-Test
Zahlenbereich
Mathematik
Gebäude <Mathematik>
Sondierung
Stichprobenfehler
Diagramm
Numerisches Modell
Verschlingung
Last
Stichprobenumfang
Subtraktion
Numerisches Modell
Wellenpaket
Mittelwert
Natürliche Zahl
Stab
Zahlenbereich
Sondierung
Numerisches Modell
Thermodynamisches System
Sortierte Logik
Raum-Zeit
Güte der Anpassung
Wurzel <Mathematik>
Feuchteleitung
Eins
Numerisches Modell
Wellenpaket
Mathematik
t-Test
Mathematik
Vektorpotenzial
Übergang
Numerisches Modell
Numerisches Modell
Prozess <Physik>
Wellenpaket
Exakter Test
Menge
Wellenpaket
Rechter Winkel
Güte der Anpassung
t-Test
Kombinatorische Gruppentheorie
Analysis
Arithmetisches Mittel
Statistik
Numerisches Modell
Mathematik
Menge
Rechter Winkel
Term
Term
Teilmenge
Numerisches Modell
Quader
Exakter Test
Sortierte Logik
Gruppenkeim
t-Test
Strom <Mathematik>
Term
Term
Prozess <Physik>
Numerisches Modell
Menge
Güte der Anpassung
t-Test
Tabelle
Analysis
Eins
Resultante
Subtraktion
Prozess <Physik>
Verschlingung
Zehn
Güte der Anpassung
t-Test
Übergang
Eins
Rechenschieber
Moment <Stochastik>
Numerisches Modell
Menge
Standardabweichung
Sortierte Logik
Minimum
Gerade
Resultante
Prozess <Physik>
Numerisches Modell
Prozess <Physik>
Endogene Variable
Mereologie
Flächeninhalt
Projektive Ebene
Reihe
Numerisches Modell
Prozess <Physik>
Verschlingung
Menge
Zahlenbereich
Körper <Physik>
Grundraum
Lie-Gruppe
Gerade
Eins
Resultante
Thermodynamisches System
Numerisches Modell
Eliminationsverfahren
Stichprobenfehler
Mathematik
Surreale Zahl
Subtraktion
Punkt
Linienelement
Mathematik
Linienelement
Güte der Anpassung
Physikalisches System
Mechanismus-Design-Theorie
Rechenschieber
Numerisches Modell
Menge
Mereologie
Projektive Ebene
Grundraum
Numerisches Modell
Teilmenge
Resultante
Numerisches Modell
Gruppenkeim
Mathematik
Vorzeichen <Mathematik>
Projektive Ebene
Ungerichteter Graph
Kontraktion <Mathematik>
Gravitationswelle
Gammafunktion
Analysis
Zentrische Streckung
Statistik
Numerisches Modell
Ortsoperator
Stab
Güte der Anpassung
Gruppenkeim
Projektive Ebene
Rechenschieber
Gerichteter Graph
Numerisches Modell
Charakteristisches Polynom
Linienelement
Rechenschieber
Projektive Ebene
Numerisches Modell
Menge
Flächeninhalt
Last
Rechenschieber
Mathematik
Ereignishorizont
Beobachtungsstudie

Metadaten

Formale Metadaten

Titel Key Note Lecture: Managing research software development - better software, better research
Serientitel The Leibniz "Mathematical Modeling and Simulation" (MMS) Days 2018
Autor Chue Hong, Neil
Mitwirkende Leibniz-Institut für Oberflächenmodifizierung e.V. (IOP)
Leibniz-Institut für Troposphärenforschung (TROPOS)
Lizenz CC-Namensnennung 3.0 Deutschland:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/35352
Herausgeber Weierstraß-Institut für Angewandte Analysis und Stochastik (WIAS), Technische Informationsbibliothek (TIB)
Erscheinungsjahr 2018
Sprache Englisch
Produktionsort Leipzig

Inhaltliche Metadaten

Fachgebiet Informatik, Mathematik
Abstract In a 2014 survey of UK research-intensive universities, 92% of researchers said they used research software and 68% said their research would be impossible without software. Yet 71% have had no formal software training, and few are ready to apply many of the things we consider as best practice in research software development. The UK's Software Sustainability Institute has worked at the forefront of activity to address these issues, providing guidance and training whilst developing ways of supporting sustainable communities of practice. My talk will challenge the workshop attendees to understand how they can work with their own communities to improve the ways that software is used and developed, including infrastructure, incentives, overcoming fears and increasing adoption, and planning for change.

Zugehöriges Material

Ähnliche Filme

Loading...
Feedback