Crazy week, nothing has gone according to plan, but I'm happy where I am at this point. I finished my paper (still needs a lot of revision, but each part is there), my project is coming to a close, and I finished my lesson plan for STAR which I think kids would have a real fun time experimenting with. For those interested...

Suggested for grades 9-12

To be completed in 1 - 2 class periods.
1.) Students will be shown real world instances where machine learning is used in everyday life (Facebook tag-a-friend using face recognition, self-driving cars, etc.). Students will break into groups and engage in a brainstorming activity to discover ways face recognition and machine learning could be used to improve their lives.

2.) The conversation will then be focused on how automated detection and the principles of Facial recognition and machine learning IS used in marine biology (using specific features unique to underwater animals to detect them among marine snow and complex backgrounds). Students will learn about 3 specific organisms (Peniagone, Scotoplanes, and Echinocrepis). Students will be asked to explore Wikipedia, google, the MBARI Deep-sea guide, and any other resource to find as many images as they can in a short 10 minute time limit. Students should focus on what makes each organism unique. Are there any species mimics? What characteristics are the most important to search for? (color, shape, texture?)

3.) Students will now be given a chance to explain these three animals as best they can as a group. What do they look like? What characteristcs stand out? What are the three most important filters (color, shape, texture) the computer should select for when trying to detect these animals? Could you use the same filter for all three or would you have to change the filter for each animal? What is more important, increased efficiency or accuracy? Why?

4.) Now that students have their 3 most important characteristics, they will use their phones or laptops to play Kahoot! And elaborate on what they've learned. They will be "trained" just like you train a neural network by being shown images of a particular species, paying special attention to their 3 characteristics. Images of the different animals will pop up on the screen and students can vote on what the animal is (Other, Peniagone, Scotoplane, Echinocrepis). Kahoot! Will automatically show which students are getting the answers right or wrong and students will be given an opportunity to justify why they picked each answer. Just like a neural network, they will get better at identifying the animals.

5.) Once students have played Kahoot! And improved their identification skills, they will be evaluated by watching an unseen video (much like a neural network would) and see how many species they can I.D. using a mock VARS system. How many did you get right? Which animals did you confuse for others (species mimicry)? Would you keep the same 3 filters (some animals are the same shape, but different color and vice versa)? Would it have been easier with more pictures (neural networks also do better with more images)?
 Topics that will be tied into the lecture period can include climate change, upwelling events, ecology, food webs, species mimicry, and evolution.
Let me know if anyone would be interested in being a guinea pig on this and we can try it out

This week I am successfully learning stress management haha. On a serious note though, things are starting to come together for my presentation and project. Since last Thursday I have been touching up my charts, graphs, figures, etc. for my results, as well as making sure I have a good grasp on what these charts are showing in terms of the question I am asking (what is the difference between the pulses). Aside from touch ups, I have been working on my presentation and paper, as well as meeting with the people in my lab to discuss their thoughts on my results and presentation. I am not as far along in my paper as I'd like to be, but the good thing is that it isn't due on the 10th so I'll have a few more days to finish it. I hope everyone completes everything on time and feels well prepared by the 10th!

I have had a pretty awesome week for my project. I have been meeting with my mentor a lot to fine tune some things in my code, which looks awesome. I have learned so much about code writing and code etiquette! The internship is coming to a close which also means that all the other things in life I have been putting off for the summer now become relevant and demanding again, which sucks, but I will get everything done.

I have started outlining my paper and writing down all my notes into one document so it will be easy to put my thoughts into words. I will be focusing my paper on spatiotemporal dominance of phytoplankton in Monterey Bay.

I haven't begun to put together a story board of my presentation yet, which is bad I know, but I'll get it done. I am trying to think about what figures are the most important and interesting, but I am biased because I think they are all interesting. There is one figure that is basically my baby and of course he will get a lot of attention but I need to spread the love and not focus all on that one.

It is Thursday and I have not packed for moving tomorrow so that will be my life this evening, however I did my confluence on time so this counts as a winning week in my book.

This week has been and will continue to be stressful. I have a 10 page (double spaced) paper due for my REU program tomorrow that could use a lot of work. Also my online class has two assignments due this weekend. I've never felt more alive.

On top of that, the sequences I received back from Stanford were of extremely low quality. In order to get some type of read that doesn't suck we have to loosen up the parameters a bit. To be honest I'm getting into the land of "what do those words mean?" when my mentor and others are talking to me. Mostly because I'm working in the terminal window and with scripts and bash etc so forth. Not my forte, but I'm learning slowly. NASA was really cool, minus the hellish drive back, shout out to my sweaty homies. Learning R steadily, but mostly just scrambling trying to write this 10 page draft, mid summer reflection, my online class stuff, and preparing for the symposium. How fun.

Finishing this week with my final draft of my product - its unbelievable how many people need to look at this before it can be approved. I understand it but it has taken so much time because of the constant back and forth.

On a different note, I wished that I had thought of this earlier in the internship but I maybe going down to Canary Row early next week to survey people. My goal is to get a sense of what the public know about the issues I have been working on such as ocean acidification and harmful algal blooms. Today, I am going to design a quick survey to get data on whether people understand what these changes are, whether they believe it is happening, and whether they believe it impacts them. I will analyze the data and incorporate it into my presentation on the 10th.

Like Alaina, I have not packed so that is going to be the bulk of today's work after work.

This was a weird week because both my mentors were gone on Monday and Tuesday, then I was gone on Wednesday, so today has been a lot of touching base to make sure we're all on the same page.

I was feeling great about my classifier this time last week, but that's because I was running it on data that was just a bunch of concatenated clicks, not raw data. I had a very unfortunate start to my Monday when I tried to apply my classifier to a clip we were hoping to use for my comparison and I got 350,000 detections, all of which were noise rather than clicks. The rest of the week has been dealing with the fallout from that realization. Turns out, that clip is actually just bad for PAMGuard (though my classifier was also really bad) so we're going to go back to the original events I had been looking at at the beginning of the internship, some files from last October. We had shied away from using that event before because we're not 100% sure it's a Cuvier's beaked whale, but at this point we're not sure ANYTHING is a Cuvier's beaked whale so we're just going to have to compare "unidentified beaked whales that are probably Cuvier's" instead of "Cuvier's beaked whales," which is a little annoying in terms of sounding cool in my presentation but in the grand scheme of my project it's fine, as long as both methods are looking at (and for) the same thing.

A really fun headache has emerged from the realization that the MARS power supply hums at 50 kHz. Why is this a problem, you ask? Because guess what else makes noise at 50 kHz! (If you answered "Cuvier's beaked whales!" then you are unfortunately correct.) I looked at some HARP data and its peak is at 51 kHz, meaning the peaks at 50 and 100 kHz I've been looking at aren't completely inaccurate, but are very strongly skewed by the fact that there's a loud hum right there from the power supply. I had been suspicious of the peaks being RIGHT at 50 kHz just because the real world does not generally stop at such exactly nice numbers, but I didn't know why until this week. I believe that if I could filter out this 50kHz hum, I could get rid of the ridiculous number of noise detections on PAMGuard, so I am finally treading into the realm of actual full-blown MATLAB signal processing, not just looking at the Welch's power spectral density estimates. Again this is a task for which I have next to no background to go off, but I'm up to the challenge.

I'm starting to feel the time crunch on getting my data. I'm not super concerned about having enough time for my paper because I have plenty of experience as a procrastination-inclined college student, but I am super concerned about having enough time to do the actual data analysis for the comparison. So hopefully all these issues will get worked out in the rest of today and tomorrow and I can do my analysis starting next week.

Also-- Tuesday was a big day in the beaked whale community! A new species was defined! Remember when I was so excited about finding out a new one had been defined in 2001? Well, that is nothing compared to a new species literally the day before yesterday! Basically everyone I know sent me the Nat Geo article about it (check my facebook wall if you're interested in the article, my friend posted it there), and one of my collaborators sent us the actual paper that was published. Originally it was thought to be a type of Baird's beaked whale, but now they think it's its own species, more related to Baird's cousin Arnoux's beaked whale, that just happened to occur sympatrically with its cousin Baird's. Crazy stuff!! Beaked whales are the absolute best, it's so cool to be in a field that is so current and evolving.

I had a pretty slow week, which probably isn't great at this point in the internship...

Ed is out all week, so I've had to be more independent than usual. Fitting algorithm/program is pretty much complete, and looking at the bending peaks of water seems to be successful (as I said last week, trying to fit the stretching peak is prohibitively difficult, there are just too many degrees of freedom when trying to fit the sum of five gaussian curves). All I have left at this point is putting together functionality to save the fits and information to a file for later easy reference. I'd like this file to include an image of the fit, the error functions for the smoothing and fit, the limits of the spectrum segment we're looking at, the smoothing parameter, the fitting parameters, and the values of the resultant fits, so I'm thinking just making a picture with all that info makes sense.

Planning on collecting some seawater spectra for the depths we missed on the flyer cruise using a temperature controlled pressure cell either today or tomorrow, so I've gone through the CTD data and collected temperature and pressure information from our three dives in 100m increments.

This project is starting to come together! Very exciting. I've been working on semantic validation, which is making sure that the code people enter isn't only grammatically correct but also makes sense. For instance: "I flew my sandwich through the Monterey Canyon," while grammatically valid, still doesn't make any sense. In terms of this project, that means making sure that any behaviors referenced actually exist, making sure variables aren't defined twice, making sure they don't use variables they haven't defined, etc. This validation is in service of an error-reporting feature that can correctly identify the location in the code where the first error is, highlight it for the user, and tell them what was wrong with what they entered. Error reporting, as it turns out, is most of the work when it comes to building anything that takes user input. As such, I'm rebuilding a lot of the parsing in order to capture more information about the code elements to provide more useful error messages. If only nobody made any mistakes, I could be done by now!

 Also, with every day comes more understanding about the source XML files, which invariably means assumptions I made early on were shortsighted and destructive. Sigh. So I'm on a bug-squashing, error-reporting, and code-restructuring spree. I haven't as much as thought about the actual report yet!... Or the seminar. Oh boy. Good luck everybody.

This week has been a little slow but pretty valuable - I now know (somewhat confidently) how to use an HPLC! In terms of purifying and characterizing my radiolarian/phaeodarian photoproteins, this is a highly useful instrument because I can fractionate my crude photoprotein extracts. In other words, it can separate sample components with different properties, i.e. isolate my photoprotein from other proteins and non-proteins in my sample. Unfortunately, 1. the machine (which is extremely expensive and requires diligent maintenance) hadn't been touched in 3 years, and 2. every brand's HPLC and associated program is different, so my only lab-mate (Manabu, Japanese visiting postdoc) who knew how to operate an HPLC had to re-learn the setup of our machine. So, Manabu and I spent the first part of this week learning how to get the HPLC to work properly, and are now playing with "model" proteins to test the instrument's accuracy and precision.

Thursdays tend to be BIG experiment days, so today I'm running another regeneration experiment. I've been having issues ridding my samples of excess calcium ions from the biolum assays I run, which seem to be interfering with the regeneration of photoprotein from apo-photoprotein. Lame.

It's amazing how quickly this internship has gone, and I feel like I have barely scraped the surface of my project's full potential. Now that we're close to the end, it seems that I have so many more questions than the handful of research-directed questions I began with. As this is the case, I am still running assays and gaining new knowledge about my photoproteins every day, which has made it difficult to begin my report - my project has no clear-cut conclusion.

That NASA trip was once again a highlight of the summer (quality of van aside). Its such an inspiring place, and I thoroughly enjoyed getting to meet with and hear from, both the staff and interns.

This week has been a train wreck as far as project work has gone. I lost about a months worth of coding due to a matlab crash, that I'm still trying to understand (as well as poor version control on my end). While heartbreaking, a lot of the code is backed up in other scripts that I've written, but I am now in a mad scramble to get back to where I was last week.

Positive steps have been made with modularizing the data so once I complete my overhaul, I can run through the analysis pretty quickly.

It's been a busy week with no time to waste, so I'm keeping this report short...

Had some hiccups this week (like leaving my external drive back in San Jose), so I've been making due pouring through additional photo and video assets for use in my web project. The glitches in ArcGIS's Cascade Story Map app became even more apparent and I have to keep reminding myself that it's still in beta. Nonetheless, I'll be clocking in some time on my website project this weekend. My goal is to have it pretty darned polished come next week. (And don't be surprised if I make rounds to have some fresh eyes critique it!)

The morning of our NASA trip, I actually debated the idea of staying behind to work on my project. I'm so glad I didn't make that mistake! My more recent visits to Moffett have only been to shop at the commissary, so it was awesome to actually check out some of the rad stuff they have going on there! Their human research program was a surprise to me. Very cool work on physiological stuff! I especially enjoyed our stop at LOIRP's "McMoons" facility and dig their technoarchaeology discoveries. I'm still blown away with the resolution of the images they've uncovered. And bonus points on the relevance and benefits of old/analog technologies!

As usually I'm writing my confluence at the very last minute!

On Monday I had a meeting with my lab where I needed to explain to all the people from my lab what I have been doing so far with a little powerpoint and I'm so happy to say that I did it well! I'm so proud of myself to be able to explain everything in English and realized that they understood what I was saying! This really helped me for the future symposium.

At this point the main experiment is done, which is awesome, but also means that I won't spend a lot of time at the lab which makes me so sad. Right know, I'm focusing at the paper and the symposium, collecting all the data and information I have from my project and trying to figure out how to put it together. I have never done a paper before about a project because I have never done a project before so this is even harder for me (and never forget that I need to do all of this in a different language, which makes this more exciting but also more difficult!). Anyways, I feel so lucky because one of my mentors is helping me a lot with this and is always willing to explain to me more about how to make the perfect paper and presentation, and also because many of the other interns has offered to help me too.
So I'm trying to decided which data is important to use for this presentation, which graphs and the best way to explain the awesome project that I have been working on during this summer.

Also NASA was amazing yesterday!!

Had a meeting yesterday with my official and unofficial mentors to get feedback on the graphs and figures I have produced so far and weed out the ones that really aren't going to be so useful for my final presentation. I have attempted a lot of different types of analyses for these datasets and it's funny but not all that surprising that the simplest of them are the most illuminating. Here I am using complex mathematical functions to produce the dominant modes of variability over the last 27 years and really the best way to show how the recent "El Niño" and "Blob" events differ from the rest of the record is a scatterplot I made in a few clicks of a mouse.

Both my presentation and paper are still in outline form but once I get my figures cleaned up and finalized I think it will be relatively easy to fill in the supporting information around them. Good luck to every one as we enter the home stretch!

Sometimes it is just better to start over from scratch. After spending the last eight weeks writing programs and setting up my sensor to run experiments on the linear table, Gene and I decided to rewrite the data logging program since the code was getting overly complicated. I also developed an entirely new strategy for filtering and processing the IMU data using a new type of filter. The good news, it was much easier and faster this time (instead of eight weeks it took eight hours) and now I have a much more streamlined process.

In addition to a cleaner version 2.0 of my code, I have added a calibration function to remove a linear offsets from the raw accelerometer data and am now running experiments to determine if the linear translation table is level and flat. Before I run the next round of experiments I will be accessing the error in the system to try and get a better estimate on the final velocity values we can expect from this test setup. We would like to measure 1 mm/sec^2, however the noise and misalignment of the stage may make this very difficult. I am sure that I will be collecting data and running matlab scripts up until the last minute of this internship, but at least the end is now in sight.

It was such a crazy week...Most of my time was spent preparing for a second experiment so lots of autoclaving and running around gathering materials. But everything went well and now I have results to work with! I guess all that is left is the presentation and paper to write, but I'm quite glad that I just need to edit my paper and presentation from the ones the REU had us do. I don't have to start from a blank page, which is a huge relief because blank pages are very intimidating...