 Thank you very much for the invitation so you know as as you were just told we've been here all week you might have seen us occupying this space and sort of frantically hacking our way through the week so so yeah this week we've been here since Monday for an event called Astro hack week this is the third time this event happens it was really it's really one of the events that are organized by the Morse and Sloan Data Science Initiative that is a cooperation between BIDS and NYU and University of Washington and it's been one of these things that really all three institutions have organized together so our organization team has come from all three and this is now the third time it happens the first time was in Seattle second time was in New York this time it was Berkeley's turn but it was really more and slow enough have made that possible both in terms of the funding but also in bringing the right people together so what is Astro hack week I'm gonna talk about this in a little more detail basically for the last five years five days around 50 of us have been in the space the people that were here represent sort of all of astronomy all of the different subfields in astronomy we've given tutorials we've given lectures and we've been hacking there are 28 projects currently in progress and the people who are now not here are somewhere in the paleontology museum to hack some of them are held in the cafe and they're frantically working because at 3.30 we're gonna have our wrap-up session where everyone presents what they've been working on so you might now sit there and ask why why why would you do this and that's a very good question the answer is sort of two falls and the first answer is that in all of science data is has been exploding for the last few years and you here at bids probably know this as well as I do here's just an example this is astronomy data archive one of many and this is just basically a plot of how the data increased between 2008 and 2014 and that's here in terabyte and so they've gone for maybe like 30 terabyte to more than that 650 terabyte in these few years and so somehow all that data out of all that data we would like to get science out of it so we need to figure out how and that's not only a problem for astronomy you know social scientists have started doing analyses with Twitter data and Twitter data seems to be growing exponentially the LHC and CERN are producing massive data sets and then one thing to point out is that it's not just data that we collect from real sources but these are climate simulations and these big simulations are basically you can consider data too because they produce massive amounts of information that you somehow need to mine and so in terms of another example from astronomy this is how astronomy used to work it used to be someone who goes to a telescope who looks through the telescope who makes a photographic plate of wanted they want to observe and then they by hand looked at that photographic plate and did their research this is not what we can do today this is an example of what's going to happen in the next few years this is the square kilometer array which is currently being built in South Africa and Australia and that square kilometer array will produce 960,000 terabytes of raw data per day so that's a lot of data that we're starting that we start to have to take into account yeah oh I'll take questions whenever these two pardon so maybe it's kind of a prosaic observation but looking at that ramp up which maybe looks a little sigmoidal now there were increased costs we're it raises the question of where's the data being stored and what are the cost burdens of that aside from the functionality that having that data gives and I mean with magnet you know like a an 8x over four years sort of thing that must mean that expenses are perhaps shifting yes so I am not an expert in the sort of structure of data management in astronomy but I think I mean this is this is an infrared science archive that's run by NASA this is all from telescope space telescopes that's to run by NASA and NASA has an growing awareness that they need to do data management and they are actually starting to do some data management for products produced by the community but yes I mean there's another project called LSS T which is a big US project the large synoptic survey telescope and their software budget alone is around 50 million dollars and a series of considerable fraction goes into a data management group that six that sits at six different institutions and over five to ten years produces the archive and produces the necessary infrastructure there's work that's being done with NCSA with the national computing facility so I think there's growing awareness also within the funding agencies that this is a problem but it is true that sort of long-term storage is something that's not entirely solved especially for the SK for example right yes yeah so that's actually not something we've talked about much this week except for a tutorial on databases but it is it is true that is one of the problems any other questions okay so what I'm trying to say is basically that the modern data that we gather is heterogeneous it comes from different sources that need to be incorporated it is usually complex the data that I work with in my daily life is actually not that big it's big enough it's small enough that you can store it on a hard disk but it's usually very complex it usually has strange strange sources of uncertainty it's usually my time series are not continuous they have gaps in them and there's all sorts of clever ways you need to try and figure out how to deal with them a lot of data is dirty especially data like Twitter data that comes from the you know from the real world but and then there's this big problem of high volume and high velocity which is current mostly called big data and so this is not only something that happens in academia that's been happening in the sort of data science industry along for a while now and the response to that was to have a sort of data science new thing called data science that incorporates different things including statistics advanced computing visualization but also domain expertise and scientists traditionally I certainly when I think back to my undergrad and to my PhD we're sort of actually only trained in a subset of that and that subset is not enough anymore to really deal with with the data that we're getting and so the point is that I think academia is really lagging behind industry and methods development and data driven research and Jake Fander plus who's at our partner institution has had two very good blog posts on that so the question that we were asking ourselves is how do we encourage researchers to learn and adopt new methodology how do we get them to even talk to people who work on the methods the other question that we were asking ourselves is within our field if you think about traditional conferences they're very seg where people normally exchange knowledge they're very segregated by subdiscipline so here's just like I just picked up the banners from the next few conferences in the next year and you know there's like an international work of spectral stellar libraries or physics of the intra cluster medium and these are extremely specialized subfields within astronomy but if all you ever do is hang out with people who work on the physics of intra cluster medium you'll never figure out what the exoplanet people are doing and maybe they're actually developing methods that help you but you would never know about it and if that problem is even worse when you think about data problems between disciplines we don't know what people in climate science are doing but they might be developing the next best tool that will help us deal with the SKA data so the second question we had was how do we get researchers to talk to each other about methods and so our general objectives for running this workshop is to teach new data science methods to researchers and to increase connectedness between them and that includes sort of fostering exchange between between disciplines it also includes built networks and lasting collaborations that people can then expand on once they go home and the sort of final final goal is to promote open science so that people that other people who cannot make it to the workshop can take advantage of this too and so our answer to this is astrohack week and I'm going to try to convince you that that's a good idea and that we can actually deliver on these goals so in order to explain to you what astrohack week is I'm actually going to start explaining by what astrohack week is not astrohack week is not a conference if I think about conferences that's sort of the picture I think of and that's a bit unfair because I actually searched for like board scientists at conference when I look for a picture but what I would like to convey here is that traditional conferences are often knowledge transfer from someone standing in the front to an audience that might listen or might not listen and there's very limited time for actual exchange during breaks or in the evening and that happens sort of accidentally it's not the main reason for the conference usually the second comment is that there is a difference what conferences do is what one of our participants called science driven communications you have someone standing in front talking about their results they don't usually talk how they got there when I prepare conference talks they just say you know here's a problem and here's this awesome result I got and here's all the exciting physics that I got out what we wanted to do is data driven communication and we wanted people to talk about here are the data sets that we have and the problems we have with those data sets and then someone else could stand up and said I actually have that same problem and I can help you solve that so if you think about how extra hack week works it's really I have hundreds of pictures of like a bunch of people standing around a laptop discussing together about a certain data problem the other thing it's not it's not a summer school even though we have tutorials summer schools are traditionally top-down so you have senior scientists on the front who will teach junior scientists something about their expertise this is exactly not what we want the hack week principle is that everyone learns from everyone and especially in a lot of ways junior people are better at adopting new methods and adopting new technologies so we try to admit both junior people who are experts at things and senior people who are open to learning from them so there is a much more level playing field in terms of teaching okay let me tell you a bit more about how it works the ingredients are kind of three things we have tutorials we have breakout sessions and we have hacking and they all sort of make up a very very mix of what people actually do at astro hack week the first time we ran tutorials we kind of decided on a bunch of topics that we thought would be interesting for the field and then run them and that worked really well so the next time we just gave people a really vague message and say we're just going to do data-driven stuff they applied anywhere which where we should really got about but we asked them what would you like to learn about and this I don't know how well you can see this but this was our astro hack week 2015 board of topics that people were interested in and we just wrote them all down and it turns out there are sort of fields that emerge and one of them is kind of inference and sort of Bayesian statistics and MCMC they're sort of data driven stuff like classification and machine learning and they're sort of exploratory data analysis and sort of big data methods and so from there we then came up with some kind of kind of program and that work that sort of bootstrapping has worked really well one thing that we learned is that actually tutorials at astro hack week are scarier for the lecturers than for the participants and this is because on almost any topic there will be experts in the room and we got a lot of feedback from our lecture saying like oh but what if I say something that's maybe not entirely correct and oh that person in the room is much more an expert at it than I am so the solution that we had to this is that we encourage a lot of peer learning because it's such a diverse group what we did was basically put people in groups and there would someone who is lecture who would give a bit of a lecture and then do exercises and during the exercises we asked the other experts in the room to help so we encourage participant to help with the teaching which worked out really well and that is a sort of formal that is the formal institute the formal learning that we instituted but that is all the sort of the tutorials are all the formal teaching that we implemented but then what we encouraged are so good but breakout sessions and they were usually short tutorials sort of 30 to 45 minutes that we encouraged participants to lead and they were usually on topics that came up during the week oftentimes they were sort of further topics on something that was considered discussed in the lectures but oftentimes they were just sort of practical things that came up so we actually have a lot of things like profiling python or licensing code and all the sort of practical things that come up and researchers daily life but that they don't necessarily learn about and one of the nice things about these tutorials is that they're usually quite on the spot and very informal so usually someone stands up like here and said I can talk about Gaussian processes for a while but I haven't prepared anything but in that sense it becomes very informal and more of a discussion than a sort of lecture and that we learned helps our helps people learn as well and then the sort of heart of astro hack week is the hacking of course this is last year we drew a sort of hack Atlas where people could write down where they were hacking so you would know where to find them but one thing I'd like to define is actually what a hack is and I'm not going to define it we asked our participants to define it last year and so I found a few which I think are very good sort of incorporate what we think hacks are so one way to define a hack is as a small project with a clear goal that has to be completable in the time that you are there you should not start you know your new research project for the next half year but instead say here's a small thing that I can try and either it works then I can build on that or it doesn't then I'm just going to abandon it but that's one of the ways that astro hack week works you need to have something that you can finish you can also it's perfectly reasonable to cobble together quick and dirty solution what you will often hear is things like oh this is a toy problem that has nothing to do with reality but if I can make this work then and that works then I can go to the more complex problem a lot of the times that involves really quick and dirty programming hacks you know code that you wouldn't want to release to other people but just making it work in the time that you can and then the third sort of definition involves simplifying a complex problem and that kind of goes back to this idea that it needs to be completable and often to make it completable you need to sort of compress it and simplify it down until you can really approach it in practice here's how hacking works every day people stand up after lunch and pitch projects that can actually also the first time we also made people stand up and say what they were good at if they weren't pitching a project they had to say what they were good at and everyone had to be good at something because everyone is but then people pitch projects there would be some time for people to discuss what they were working on and this year we had lots of small hacks there was lots of movement between groups and then people hack and that these hacks can take an hour they can take four days depending on the project at the end of each day people put present their results their results are it has to be something tangible it can be a plot it can be a blog post it can be you know a notebook anything but there has to be something tangible that comes out of a hack that tangible thing could also be a plot that says what I tried really didn't work I failed on this but that's okay and then they learned something from it and that's good too so just to show you maybe that's a work okay so here's the example we've been doing everything on hackpad and that is our sort of central open source of everything that happens and anyone will add their hacks and you can see there's people working on this right now and they are sort of bunch of active projects with links to like github with links to to ipython notebook with links to explanations and so anyone of us can sort of go there and read and that's where we also go at the end of the day to help people to look at hacks that happened it's at this link I think my slides will be online afterwards right maybe yeah I can put them online and then all the links will be there okay so here's just this is one of the few pictures where I have people smiling it turns out no it turns out people are not actually unhappy they're just extremely concentrated there's usually like a bunch of people sitting around a computer going like trying to figure out something really hard but this is just really how we work it's a bunch of people sitting around a laptop really really getting things done in practice one way we encourage open science is by just asking people to do everything openly that happens at astro hack week all the descriptions at the hackpad are officially you know can be seen we encourage people to put stiff in git and on github just to be ever really open and so that's that's one of the basically the structure is the thing that we can steer the other thing that we can steer that is really crucial to astro hack week and then we can only steer to a certain extent are our participants because they're a huge part that makes up makes up astro hack week and the decision that we made for astro hack week early on that we decided that diversity promotes excellence and that we wanted to be a really that we wanted astro hack to be a really diverse group and when I mean diversity I mean diversity along of along many different axes and one of them is that we want a broad diverse diverse group between binnett Guinness and experts that can be beginners and experts in statistics and machine learning in stellar evolution in anything really but it's again different from a summer school where you have sort of experts who teach and then everyone else is a novice at astro hack week we're trying to get people from all levels to be involved this also includes sort of diversity in academic seniority we got under everything from undergrads from senior professors and again in a lot of ways senior people can learn from junior people as well especially you know junior people person stands up and say hey I'm going to do a git tutorial and most of the senior people I know actually don't use git so sometimes they go like hey I can finally learn this git thing my phd student keeps talking about one of our reasons for doing this was foster exchange between topics and so we also try to admit people from different disciplines within astronomy and usually we actually have people from outside astronomy as well there's usually a couple of computer scientists there there's sometimes people from industry there's sometimes last year we had I think a chemical engineer the year before we had someone who works on sort of geospatial map data and somehow these people always also bring extremely valuable experiences in the group and then we also care a lot about diversity and background and with that I mean sort of you know diversity in terms of gender in ethnicity and racial background because again we think it's important for the group to have that so astro hack week is really as only as successful as its participants so one question that we ask is how do we get a good mix of participants where mix means you know we want to diverse group with different backgrounds and academic sub-disciplines and everything from beginners to experts and all of our interesting interesting things that people could be beginners to experts to and also everything from undergrads to to academic seniority and when I organized astro hack week in New York last year I sat down and said these are my goals when I find my participants I wanted to optimize the mix for all of these diverse things I wanted to reduce my own bias sort of unconscious biases I also kind of wanted it to be transparent and I wanted to be accountable for how we do that and so that was basically me last year when we got 168 applications for 50 spots and I was like what what do I do with this and I was very fortunate because my office mate is a computer scientist and he said you're a data scientist you should be data science with this and so the first thing that we realized that once you have a group of people that you're happy with coming and that group is still larger than the spots that you have what you have is a complex optimization problem and it turns out computers are really a lot better at complex optimizations than humans are so we decided to let a computer do it and so it's got sort of two and a half steps one is find all in your in your input set find all your your participants that you think are appropriate to have at the workshop and we asked a few questions like why do you want to come to extra hack week and if someone had given a completely crazy answer we had said well maybe maybe that person isn't actually serious about coming then we also pre-selected a few experienced hackers because you do need some seeds for people who can spark you know hacks and who know what they're doing and then we used an algorithm to break ties between the rest of the people conditioned and our what we thought was our ideal mix of participants and so here's for an example of one of our cat several categories one was statistics experience and that goes sort of from little to none to expert and we asked people to rate themselves and here you can see the blue is the sort of fractions of our input data set and then the black lines are the targets that we set the algorithms and then in blue and green is what actually comes out and so we decided we wanted a bunch of experts we wanted a bunch of people who knew nothing and then we wanted most of them in between but there was a conscious choice that we kind of made and as you can see the algorithm can't doesn't always completely manage to hit the goals and that it's because it's optimizing over a large range of different categories so in the grand scheme of things depending on imbalanced your input data set is it can't always do that here's here's the other one of the ones I care about we asked people very much care very much about we asked people whether they considered themselves a minority in the field in terms of gender identity and we also asked them whether they consider themselves a minority in terms of racial or ethnic identity and one thing that we learned previously is that representation matters a lot we heard a lot of from our especially female participants that they said they felt much more comfortable at the workshop because there was a significant fraction of women there so that was also conscious choice that we made to optimize beyond the ratios of the input data set because we thought it would be good for our workshop this code is actually also an astro hack week project it's online if you're interested for your own own workshop I really I really care a lot about the sort of selection problem for participants and I would love if people wanted to talk about that more because we're still sort of optimizing our approach and approving every time I do this okay so I talked a lot about how we make astro hack week work but you might still say well but does it actually work does it what was supposed to do so we run a survey every every year and I'm not I'm not a social scientist so I'm not really qualified to evaluate this in the sort of quantitative way but I can give you a few things I can show you a few things that we learned from the survey so one thing we asked them is what did you learn at astro hack week one thing that people said they learn is not to be too afraid of Bayesian methods which I thought was a great success they learned sort of some machine learning and where the resources are and there were sort of corrected misunderstandings about things they already knew and that was the thing I would really wanted really to to see that they thought they learned some data science method they learned machine learning and especially they learned where to find resources that they can can go to ourselves and that was really one of our goals they also said they learned about new tools they learned about practices they improved their programming skills they learned about profiling and commenting code they learned about team coding one year we enforced pair coding for a day and so these are all the practical things that we wanted people to take away all the things that would make their make would make their day-to-day life better and make their computational results better and improve open science and then there was sort of more social things one of the one one person answered that the main thing they took away from astro hack week was that they're not alone and they went on to explain that they're the only person in their department who does sort of data intensive research and that person said they felt really alone and they didn't think anyone else was doing this and coming to astro hack week they like there's this whole group of people and they came away with many connections that helped them and then there were sort of specific specific interactions where they said Brendan Brewer is a statistician we invited and they sort of drove a new project forward by quite a lot just by meeting that person astro hack week these are all sort of anecdotal we also asked them questions that relate to to the more slow and goals and one of them is sort of one of the big goals is to learn about data science career so we just asked them do you believe that astro aga was useful for your future career and of the 28 people who responded basically all of them said to some extent yes except for two who said they don't know so similar we have similar results for the questions that they think that the skill are applicable outside academia which is also important in terms of making sure researchers might transition to industry from academia we asked them do you feel that the things they learned improve their day-to-day research and most of them did which I was really pleased by because we really wanted everything to be applicable and then we also asked them whether they felt that it made them more comfortable to do open science and a lot of people there were a lot of discussions about how to do open science how to write papers out in the open and most of them came away with a feeling that they're happier about that do I do I have time okay if you if you don't mind I'm gonna show you just a couple of projects to give you a range of ideas of what the kind of stuff about come up okay so here's one they're all really sort of astronomy so I'm not going to go into much of the detail but here's a project that I picked for one main reason and the main reason is that the paper is written by a group from Chile and a statistician from New Zealand and so these are two groups that are very geographic and very different geographical locations they come from different backgrounds and work in different institutions one is an astronomy department the other one is a statistic statistics department but they managed to get together at astro hack week and start a project that ended up in a paper a few months after astro hack week so and this was a month this was a project that used some very complex sampling algorithms and very complicated statistical methods to find planets in in what's called radial velocity data from stars and that's generally a hard problem and they managed to to get a good result out of that as a and you know that that collaboration started astro hack week and persisted beyond that this is a project from this year not everything is in astronomy so that was a project where they looked at the fraction microscopy of molecules and one thing that we discussed a lot at astro hack week this year was optimization methods including stochastic gradient descent and there was this group that said we want to learn more about stochastic gradient descent and we want to use it and I think I think this is gonna yeah this is a video and then the video shows you how how the model sort of slowly converges on the molecular structure which is quite similar to the one they put in so that's a project that has nothing to do with astronomy but uses expertise that people built up from astronomy to solve that kind of problem there's sort of software projects SN Cosmo is a big software code that deals with exploding stars and they were sort of models that for a long time should have been implemented but weren't and at astro hack week a couple of people sat down and said that's gonna be my hack we're just gonna implement this new feature in the software and that happened there were a couple of people who decided they wanted to learn more about machine learning and they picked a project they are interested in in this case these are sort of large-scale structures in the universe so every point in here is a galaxy and then they decided they really wanted to find filaments in there and sort of clusters in there and they use machine learning and here is an image of the filaments that they found the filament confidence and in black you can see these are all what the machine learning algorithms are things as filaments and these is what the machine learning algorithm things as sort of clusters over densities of galaxies and so they started out not really knowing anything about machine learning and by the end of the week they had this sort of working example and people write tutorials today we had a tutorial about auto differentiation and that's something where someone said I would like to learn auto differentiation and that was in the experiment this year where we said you learn about us you group learn about auto differentiation and by the end of the week you're going to give a breakout on this and that's a kind of scary thing for people to do but it was useful because there was expertise in the room that people could immediately use once they had a point even in their breakout session where someone asked a question they couldn't directly answer so that worked really well there's also some projects that have kind of nothing to do with astronomy somehow we ended up talking a lot about color maps this week and so one of the hacks was a few people who did this project called urban goggles where they make custom color maps based on queries about cities on flicker so for example they gave it a query string Manhattan at night and then it produced a color map for your plots based on that or they gave it San Francisco and you know San Francisco has a different color map so there's these kind of projects usually that people also do to teach themselves stuff and what else do I have this was then one that was born out of a needs because we have a lot of people I straight went to learn something and we have a lot of expertise at Esther hack week and it's not always clear how to pair up people so these people asked asked the participants what do you want to learn about and what you're expert on and then they just build a graph model for matching up people and so every person you know this person wants to learn about grab probabilistic graphical models this person knows about graph probabilistic graphical models so there is an error between them and they did now the same I don't have a profit but they did the same for breakout session where they say these people want to learn these people have the expertise here's how you should do your breakout session and then the final one I have is they built a Twitter bot based on David Hogs David Hogs a faculty member at NYU based on his blog and his his Twitter account and they just wanted to learn how natural language processing and deep learning worked so they built a Twitter bot that would just respond to you if you tweeted it it's unfortunately no longer active but you know they said they learned a lot from it and it was a kind of fun thing that they wouldn't do in their daily research life even though they're going to take carry this knowledge back with them and they might be useful so as the last thing I have a few lessons that we've learned so far and we keep learning one thing that we learned is that participants come to Astro hack week with vastly different expectations and goals and that's a function of the diversity of people who come and that's both a good thing because anyone can get in something interesting and useful out of Astro hack week we have people who come with the expressive goal to learn about how hacking works and how hack how to organize hack weeks we have for example undergrads often come to learn about methods or even PhD students learn about sort of machine learning and Bayesian statistic all the thing they think will be useful in their research it's also a problem because it makes it hard to evaluate whether the workshop was successful because if everyone comes with different goals then it's hard to find common metrics to evaluate whether it worked the other thing that we learned is that student travel support is absolutely crucial we had I had people come to me and say I am glad you could fund me because my supervisor would not pay for this so there are you know cases where supervisor said oh just go to a traditional conference this isn't really useful also if you do want to have undergraduate students there's almost never any money for undergrads to travel and our the undergrads that came here have always been incredibly involved and have been incredibly made incredibly valuable contributions so we really encourage to have them there this relates directly to the more slow working group the space where you are matters in various ways it needs to facilitate hacking and in in that sense we're extremely grateful to bids because this space was perfect because it had these tables they can easily reconfigure all the chairs move and so you can quickly move around and say my group needs a table and they go just go pull it to the side and do it there's a bit of opportunity for people to sort of have breakout sessions while the rest of them works and that's been really important too but so space is really important to us one thing that we continue to address is the imposter syndrome is everyone here know what the imposter syndrome is the imposter syndrome is basically the feeling that you don't belong into the group where you currently are that everyone around you is much better at everything and you is and you're an imposter you don't belong here and at some point someone will find out that you don't belong here and will throw you out and it turns out that's that happens we learned that already during our first astro hack week we that happens to basically everyone we told people this year it happens to the undergrads it also happens to the senior professors I said earlier that our lectures you know every lecture I talked to last year before they talk they were like oh my goodness there's like this person in the room who's much better at this than I do and what if I say something wrong and oh my goodness and that is also kind of an expression of the imposter syndrome so one of the things what we've done is just explicitly address it at the beginning and tell people you will feel that way and you will feel that way because everyone around you is an expert at something they'll be an expert at different things and that means that it can feel like everyone around you is an expert at more expert at everything than you are and I think having people having making people have that in their minds help helps address it but if anyone has ideas about how to do that better I would love to discuss that with you too and as I said earlier especially for minority participants just having representation helps a lot as well in practical terms food and coffee are absolutely crucial especially the coffee and one important thing that we learned is that you want to have it in house you don't want to have people go and fracture up because they will fracture up along the groups that already know each other so you really want people to stay in there which means which means getting in catering the evenings are important to us too it turns out it is really useful to know a bar nearby that has big tables wi-fi and beer and I can guarantee you that people will just go and keep hacking that happened every time this this so far and it's here I just followed people I don't remember so this year's organizer yeah yeah so Kyle our organizer was the one who sort of scoped out places it is it is a good thing to know yeah Kyle Kyle would be the person to ask that is true yeah we did we try to get we try to get sort of usually have one afternoon or one evening somewhere else and we were quite lucky that github offered us their space for the afternoon on wednesday and so on wednesday we sort of hacked at github and it was it was incredibly exciting for everyone because we we do all our work on github and so just to be there was very cool it's also just a very good space for hacking so a lot of it got done there's also a lot of things we actually don't know and so we still don't have a good idea of how we can measure success we ask people whether they think they were successful and learn something and sort of hit the metrics but measuring this quantitatively is actually quite difficult and that's sort of an unsolved problem sort of one question that I keep grappling with is what's the sort of ideal mix of participants in terms of their different attributes that we're interested in we made a choice we made certain choices for this it is not clear to us that those are the best or ideal choices so we continue to think about that too another question is what's the right balance between learning and project work the way we did it was we had lectures or tutorials in the mornings and then the afternoon's free for hacking and for breakout sessions and this year for the first time we did it that we had only four days of lectures instead of five and the final day was just hacking we'll see how that works out and then one question that I keep also thinking about is how we can mitigate imposter syndrome how can we make shy people stand up and talk as well because we don't want astro hack week to be dominated by a by a few people who are very confident in their abilities because that again increases intimidation for other people so let me finish with asking you should you run a hack week and I think here are the reasons why you might want to run a hack week you might want to run a hack week if you have more data then you won't know what to do with or and or your data is very complex um you should also run it if you think you might need better or new methods that you need to get and maybe from other fields um if you think this kind of collaboration and networking is important beyond what you get done and sort of um the breaks at traditional conferences astro hack week is a really good way or a hack week is a really good way to make that happen because there is this space and there's this time for people to start these collaborations and sort of get projects off the ground um we think it's also been really good at promoting open science and we keep hoping that people will carry that back into their home institutions again how to measure that is kind of difficult so conclusion yes you should run a hack week um there is no it's not not just astro hack week anymore actually next week there will be neuro hack week in Seattle um and that was sort of originally based off astro hack week for neuroscientists and then there will be a geo hack week in november as well and both of these are at the east science center where astro hack week got started as well so they've been very active in this um i've put some resources on the slide if you're interested our website is on there we have a github organization for astro hack week where there's all our materials and actually most of our organization happens in github issues so if you want to you know follow along with some of our discussions that we've had about how to make this successful and how to invite speakers um that's all on there um there's our hack pad if you want to look at our hacks i've written a bunch of blog posts about how to organize astro hack week with all my experiences from last year and i continue to keep writing blog posts as i learn more um yeah the the tool that we wrote is on there as well and then there was an incredibly insightful blog post earlier that was called the horror of hack days by someone who really hates hack days for many of the reasons that we think about the imposter syndrome and we tried very much this this um this time to address that type of person and make it easier for them no no no that was someone just from tech but this got linked somewhere on twitter and i saw it and i thought oh my goodness i bet there's people like that among us as well and i bet there's people who might not want to come to astro hack week because they might feel like that and so we've been thinking about how to address this and how to make astro hack week welcoming to people who are more maybe introverted or something like that um and that's basically all i have to say thank you for giving me up the opportunity this has been a big group effort kyle barberry has been the main organizer here and i'm sure if you have specific questions he'd be happy to answer too the organization team in general has been kyle barberry, phil marshal, david hogg, jake funderplass and me we've been sort of doing this for the last three years um i think fernando was on the organizing team the first time too um and then brian mcphee has been helping me in um a lot with with the participant selection part of it um so thank you again and i'll take any questions you might have