πππ§π£π’ πππ
@KenjiEricLee
Followers
1K
Following
6K
Media
182
Statuses
3K
Find me at @ kenjilee on π¦sky. Ph.D. student in the Chand Lab @BU_Tweets. @NIH_NINDS F31 Fellow. @cientificolatin mentor.
Boston, MA
Joined February 2018
Tweeprint for those interested! We developed a method, PhysMAP, to identify cell types directly from electrophysiology and show that it interpretably reveals excitatory and inhibitory cell types (PV+ and SST+) of different layers in mouse cortex. 1/8
2
15
69
Flight attendant on the intercom: βthank you for handing me all your poster tubes; there were 19 of them. You can pick them up at the front when you departβ
48
182
5K
@francois18 @cosmojellyfish Highly highly recommend the podcast Offshore by @CivilBeat if you want to delve into a nuanced, contemporary view of Hawaiβi.
7
39
761
Congrats to the GRFP winners! unfortunately for most, today they received disappointment. I was rejected from GRFP/CSGF/NDSEG likely because my GPA was awful. Here are the scores for my F31, years later. 1 = best (1 to 9). If youβre good, theyβll see it eventually; keep at it.
2
8
229
From Bidenβs Campaign Pledge on immigration, βBiden believes that foreign graduates of a U.S. doctoral program should be given a green card with their degreeβ.
We need to staple a green card to every PhD diploma earned by an international student. Trump's executive orders have hurt the reputation of the US for its openness to international scholars. We need to more than repair this calamity and canceling these orders won't suffice.
5
36
221
We are pleased to share our new study from the Chand lab at BU that uses UMAP and graph clustering on single-unit waveforms to identify candidate cell types in monkeys and their potential role in perceptual decision-making.
Non-linear Dimensionality Reduction on Extracellular Waveforms Reveals Physiological, Functional, and Laminar Diversity in Premotor . #biorxiv_neursci.
6
41
184
@TonyZador The difference between the U.S. and other nations is absurd. Posting this graph in case some havenβt seen it before.
13
39
146
At long last! First in vivo Neuropixels recording.
4
2
136
Chand Labβs first neuron!! β‘οΈβ‘οΈπ§
2
7
135
Super excited to announce our manuscript out in @eLife! We developed an analysis, WaveMAP, combining graph clustering, UMAP, and interpretable ML. When applied to extracellular spikes, it uncovers cell type diversity hidden by feature-based methods 1/14.
4
18
126
Sometimes neurons will tell you *exactly* what theyβre up to, no analysis needed :)
6
4
101
Delighted to see that grades will no longer be considered! If youβve done enough to get into a PhD program, they shouldnβt be taken into account given theyβre no predictive of grad school success.
π’NIH is pleased to announce revisions to the fellowship application and review process for grant submission deadlines on or after 1/25/25. The goal is to improve the chances that the most promising fellowship candidates will be consistently identified.
4
5
87
It's a bit of a strong statement and I'm confused as to why the non-preservation of ambient distances is held up as some knock against UMAP when the UMAP paper explicitly eschews this. It doesn't preserve ambient distances because it's not supposed to!
It's time to stop making t-SNE & UMAP plots. In a new preprint w/ Tara Chari we show that while they display some correlation with the underlying high-dimension data, they don't preserve local or global structure & are misleading. They're also arbitrary.π§΅
2
18
87
It was nice to finally go viral at least once before the meteor hit. Follow me at @kenji@fediscience.org I guess! For the interested, this was the Delta red eye out of San Diego to Boston on the last day of @SfNtweets. The attendant was very happy to help despite it being 10 pm!.
0
0
72
Really really important work by Max Shinn to shine a light on how PCA can show oscillatory components from noise. A definite must-read for anyone in neuroscience. Phantom oscillations in principal component analysis | PNAS
2
18
71
BU is now threatening to dock the pay of striking grad students. My standard of living has actively decreased since starting my PhD but your support would go a long way towards helping us get back to a more stable living situation. Thank you and please RT!.
1
34
63
This seems to me like itβs just what happens when you apply PCA to noise: the top PCs resemble harmonics. Same issue is that one high-profile fMRI paper that showed harmonics over brain surface.
Cortical activity is made of rhythmic patterns of various frequencies. It resonates when stimulated at specific frequencies. Interestingly, resonant standing waves in the gamma band correspond to a spatial pattern of around 1mm (~ size of a functional cortical module?)
7
7
69
If you're headed to Cosyne this week, stop by and see what our lab has been working on! (also lmk if you're interested in hitting the local climbing gym!)
2
7
58
Finally here!! We were too scared to open the real box for a photo. Thank you @imec_int for the first sale of 4.5 cm NHP Neuropixels!
1
2
60
New paper from Chand lab! We use multimodal graph combination to identify cell types in several areas of mouse ctx from ephys alone. Verified via optotagging. Building towards a future with cell types in every recording. Tweeprint when im not at Cosyne!.
5
9
57
Why is this framed as βprofessors vs. studentsβ? As if, students are selfishly eating up lab costs and not instead giving up their 20βs. If we want to tell students the βreal costsβ we need to be transparent about graduate outcomes and talk about lost income opportunity costs.
4
5
55
@gvanrossum Sapiens and Guns, Germs, and Steel sound great and all but theyre really just narratives spun out of scant to no evidence. Anthropologists sort of hate both books. is made of academics and they have a great reading list on the sidebar.
0
2
51
From Cosyneβs opening session: βPillow lab won Cosyne this year as the only lab with 5 abstracts accepted!β Me in the back: π€π« .
If you're headed to Cosyne this week, stop by and see what our lab has been working on! (also lmk if you're interested in hitting the local climbing gym!)
1
3
51
Waiting at the terminal post-SfN for a red eye to Boston and seeing people working on their Cosyne abstracts is stressing me the fuck out.
1
0
50
Maybe not controversial but if youβre interested in neuroscience, donβt apply to neuro PhD programs. Theyβre too ridiculously selective; go for affiliated programs. Also, no one cares about your double major or classes, volunteer in a lab 20 hours a week and youβre set.
Which of your STEM education related opinions would get you in this position?
4
2
48
@doristsao Welp. Just got into grad school and was so thankful to have my poor undergrad grades behind me but now I guess I can go back to worrying for the next 10 years.
1
1
48
@Metabo_Dave Iβm one of the students who responded and these numbers are no surprise at all. I took way more issue that he seems to use it as an argument that studentβs shouldnβt advocate for greater stipends because it increases costs. Why not put the pressure on admins or founders instead?.
2
1
46
@benoitbruneau @drbstewart Neither densities nor original Euclidean distances are preserved under this nonlinear dimensionality reduction. Therefore, velocity and distributional overlap donβt make sense to compute.
3
1
48
@francois18 @cosmojellyfish @CivilBeat This is the most popular Iβve ever been! Check out my soundcloud!
1
3
43
Thanks to @univpugetsound for featuring me! Big thanks also to the PI's (@SiddBits, JΓ©rΓ΄me Lecoq, and @chandmuse) and institutions (@HBAEagles, @univpugetsound, @UW, @AllenInstitute, @BU_Tweets) who've supported/believed in me! I should've put a better plot on the computer . .
Logger Kenji Lee β15 is intent on improving the treatment of neurological disorders. Today, heβs a doctoral student at @BU_Tweets, teaching computers to understand the brainβs commands so he can build better prosthetics: INNOVATE #LikeALogger. #PSsuccess
0
5
42
Time to play some βDeepLabCutβ and some βKilosort3β. π¬
4
1
42
Love studying at coffeeshops in Boston/Cambridge: every laptop has up at least one of (a) Mendeley, (b) a manuscript in progress, or (c) a python/Matlab window.
i am at a restaurant and there is a man showing a woman at the bar a full blown PowerPoint of his research on transmembrane proteins?????? this is TOO BOSTON.
3
0
45
We're excited to put out this protocol (now in @STARProtocols) on how to use WaveMAP on your waveform dataset! It's been very bracing to see others use our tool and hope this manuscript will provide an entry point for others hoping to use it.
1
5
40
@Meliseymo Yup. Imo assessing the βresearch potentialβ of late UG/early grad students off of a 2 page proposal is futile and ends up reinforcing existing inequities (I.e. GPA+prestigious undergrad+mentors who can help write good proposals). Itβs a subsidy for already wealthy top schools!.
1
3
36
@KateWassum (Incentive) structures in academia: how grants are evaluated; peer review process; how tenure packages are evaluated; what do hiring searches look for/how to appeal to industry; responsibilities/expectations of a PI; subfield-specific ecosystem of labs, conferences, and journals.
2
0
32
Our updated manuscript with new DLPFC data that is consistent with predictions from our multi-area RNN model is now out and being reviewed at eLife! Congrats to @MichaelKleinman, PI's @JonathanCKao & Chand Chandrasekaran, @TianWang0318, @ColeeCarr, and others!.
A cortical information bottleneck during decision-making #bioRxiv.
3
7
36
I mean, for a PhD advisor, I would argue the mentor is one of the only things to consider and the research topic is almost entirely irrelevant.
1
1
37
Lisbon is looking amazing but bring your walking shoes! I was not ready for these cobblestone hills but I am ready for #Cosyne2024.
1
0
33
@ATinyGreenCell @kaia_mattioli Ugh except I have literally done this on a date with a girl at a bar in Cambridge at the aptly named Miracle of Science. In all fairness, it was on my phone and she asked. She also showed me her manuscript on arXiv. There was no second date.
2
1
30
Come see work from the Chand lab (co-led by @TianWang0318 and I) on how the computational geometry of multiple tasks is preserved during decision making across macaque DLPFC and RNNs! Poster 2-016
1
2
30
Thrilled to announce Iβll be attending Boston University in the Fall for a PhD in neuro! A big thank you to all who helped namely @matchings , @DrJigsaww , @adriana_et_al , @lucylai_ , @SiddBits , and others not on Twitter! And of course, @chandmuse for welcoming me to his lab!
2
0
27
Just published from the Chand lab! Congrats to all the co-authors on their hard work!!! 22 supplementary figures π.
New paper from the Chand lab!. Initial conditions combine with sensory evidence to induce decision-related dynamics in premotor cortex.
0
1
26
@pauldauenhauer Somewhat uncharitable but this easily reads as "You PhD students should feel lucky you're getting paid at all!".
2
0
26
@KateWassum Also all the computer things that are used in research but aren't taught in CS for life sciences classes such as linux, terminals, VCS, documentation, data standards (NWB), code reproducibility, real-time programming (NI or Simulink). Pretty much
1
1
23
@heysciencesam I use an iPhone app called CVSimulator to make sure all my figures are still clear for a variety of types of color-blindness.
2
2
24
Is it possible for an animal to be *too* good at a task? Not session averaged! Perfect curves over only 300 trials! He focuses better than I do .
2
0
25
@matsci_mikayla I have heard of faculty negotiating for one in their offer package lol. Explaining that if they want you to attend hoodings, they better be the ones to buy it for you.
0
0
27
I was inspired by Krishnaβs work as a freshman; some kids knew every football player and the teams they played on, I had scientist genealogies. I remember deeply wanting to be on this branch and, a decade later, there I am! Grateful for Krishna and grateful to be a part of things.
Who was @shenoystanford?.The change in conversation from neuron to population,.the development of dynamics in neuroscience,.the advance of BMI to help humans with profound illness. But mentorship and care for his neuroscience family, as he would say, that was deepest Krishna.
1
0
25
Back home for break and decided to explore the park behind my house for the first time.
2
2
23
Boston people know this probability is, in fact, rather high.
Your chances of being hunted by a turkey are low, but never zero.
3
3
26
@Metabo_Dave I get that it's hard to be a PI and have to watch the costs of doing science increase with no commensurate increase in funding but it's also hard to be a student experiencing the costs-of-living increases with no stipend adjustment. I take issue with valuing science over people.
0
0
21
I thought this way esp. as an UG/grad student in app. math. This changed after working in exp. labs for 7 years. I was ignorant about how complicated it all was until I started collecting data myself. I thought "complex math" would make sense of "complex data". I was very wrong.
Last week, a prominent computational neuroscientist suggested to me that in 20 years we may look back and realize we already had all the experimental data needed to understand computation in the brain in 2022, but institutional failures held us back. Do others agree?.
1
0
24
Awesome to see the Brain Observatory paper now out there! This was what I worked on for the past three years and the team effort of dozens of individuals at the Allen Institute. Congrats to all involved!.
0
4
22
Once again beating a dead horse and saying taking a few years off before a PhD was the best decision ever. The Allen Institute matched contributions at 50% *up to the federal limit*. I put away $50k after tax and am halfway done for retirement after only 3 years!.
If you fill your Roth IRA for 10 years at the beginning of your career, and get 7% real returns, and never contribute another dime during 30 additional working years, we're talking $670k at retirement. 10 years in your 20s is worth so much more than 10 years in your 30s.
1
2
22
Cool results! Very gratifying to see people using Brain Observatory data; makes all those countless hours in the dark worth it π₯².
My paper with @jpillowtime analyzing the Allen brain observatory Ca2+ imaging dataset is finally out! #tweetprint to follow.
0
1
21
Please retweet! Come see what weβve been working on in my first year at our labβs *first* poster, βRevealing cell types in vivo via dimensionality reduction and graph clustering of spike waveformsβ (P380 @ 3:00 PM EST on 1/11). #SfNConnectome21
4
6
20
Yes this is why I worked for four years before a PhD through with an employer with a great match (50% to federal limit on a Roth 401k). Investing a modest amount of money that early is enough to retire on without even putting anything else in! So many students donβt realize this.
1/ Most institutions do not offer 401k/403b plans to graduate students. This results in considerable loss to an already underpaid and undervalued workforce.
1
0
22
@WhySharksMatter Itβs because we have to ask why he feels the need to address this to grad students? If itβs a PSA, why not direct this to postdocs on the market? He tells on himself here
2
0
21
Mention in passing to grandma that I βwrote a paper for schoolβ and now this
2
0
20
@doubleshelixpod @jpillowtime This was helpful but I think many of the questions were a bit too . pointed? I donβt think a grad student is likely to say what frustrates them about their PI outright; maybe instead asking them what sorts of qualities a grad student needs to succeed in that lab?.
2
0
19
Iβd never met someone with a PhD until I got to college. I donβt think Iβm gonna make it to faculty yβall.
5
1
21
To summarize, it's unfair to criticize UMAP for something that it explicitly states it does not do (ambient distance preservation). In fact, this lack of ambient dist. preservation it's key to why UMAP works! Overall, this is a good awareness piece even if I disagree on the tone.
2
1
18
Come check out work from Chand lab (and collaborators @MichaelKleinman @JonathanCKao) at our first SfN! We show, through neurophysiology and modeling, how PFC areas coordinate during decision-making. Also how we find cell types using Neuropixels/WaveMAP in macaque V1. Please RT!
0
3
15
One thing I didnβt expect about monkey training/shaping was how individualized motivating them is: time of day, juice type, snack breaks, nap breaks, timeout length, secondary rewards, stochastic rewards, whether they like even the experimenter etc. very fun & very challenging!.
Mouse training is an art form, but there is one tenet I stand by: they naturally go from being hypermotivated to not motivated, and while you may control this to an extent by adjusting reward-to-punishment ratio, it's best to sort trials into states (e.g. Ashwood et al. 2022).
3
1
19
So I usually go a little stir crazy on my plane rides home but this time I had my F31 revision to work on; this time, more than anything, I was annoyed the flight had ended so soon! Even the couple across the aisle commented βyou stared at a stack of papers the whole 11 hoursβ.
1
0
17
A lesson that needs to be taught to all grad students EARLY not just to understand how much they could be getting but also to understand the long-term opportunity costs of academia. Especially for quantitative types, this number is in the *millions* of dollars.
PhDs must be recruiters' favorites for salary negotiations - we are so used to being underpaid and underappreciated and don't know our market worth at all.
1
1
14
Utterly floored that our lab's method was used in analyzing the first (?) human neuropixels recordings. Kudos to the team! Sincerely, I'm honored to have been able to be some small part of this historic effort. Another point on the board for UMAP!.
Really proud to share our preprint on the work, and lessons learned, in using Neuropixels to understand the human brain with an incredible team. @EricMTrautmann @sergeydoestweet @MGHNeurosurg @MGHNeurology @shenoystanford @RMarkRichardson @MGH_CNTR
1
0
19
@pauldauenhauer And that's great for you! However, it is also important to realize that, as a mentor, not all students have that luxury. In fact, >90% don't. I too am grateful to be in a PhD program but gratitude doesn't keep a roof over my head.
1
0
14
This drained me emotionally, in the best of ways (like listening to Mahler). Besides being an incisive critique of the rot in Academia, both structural and specific, it's a sublime piece of writing. Maybe @theorygurl can write a take-down of life science?.
1
1
18
UMAP got me my first first-authored manuscript in my first year! I canβt overstate how useful it would be for so many neuroscientists but hasnβt yet for no better reason than that they donβt understand it. Read this thread (or DM me) if you donβt!.
The latest version of umap-learn is now out. Version 0.5 includes some major new features, including ParametricUMAP, DensMAP, AlignedUMAP, model composition, and model updating. Thank you to everyone who contributed! 1/14
1
0
18
I did *not* move all the way to Boston just for my family in Hawaii to get the vaccine before me! π‘ enjoy huffing that raw air.
0
0
18
Ugh why is there always a line outside of Trader Joeβs
0
0
17
Anyone else headed to #Cosyne2024? I, uh, got my poster rejected but Lisbon is too good to pass up!.
3
0
15
@sabrinastronomy 1000%. I took 4 years post-bacc to build a nest egg and pay for loans and Iβm relatively privileged financially (parents paid most of my tuition). People shaming others for going into industry have generational privilege and it shows.
2
0
15
Quite frankly, I find this graph disgustingly racist. 1) Selectively disaggregating Asians and only showing those that are high-income but omitting/erasing others that are less so like Hmong, Cambodians, etc.
3
0
17
@EpiEllie Letβs just say that after seeing Brookline public health fit a Gaussian to times series data (to show COVID was no longer a problem), I donβt exactly have a lot of trust in their data-driven decision-making. Will leave an email though!
5
0
15
For a very long time Iβve been asking people for a guide to extracellular Ephys and the best reading out there was always βthe Axon Guideβ. Hereβs what I wish I had years ago!.
Quick ephys test: . β‘οΈ2 identical devices start acquiring at the exact same time & frequency. Why are they not synchronized? . β‘οΈWhen should you not connect GND and REF? . β‘οΈWhat happens inside a headstage?. Drawing a blank or just want to know more? You're in luck! . (1/6)
2
3
15
Thanks to Alex and the CCN for an amazing time!!! The institute and people (and food) were π€©. Not sure why I decided to mirror @emdupre_βs pose lmao.
Just wrapped up the @FlatironCCN Junior Theoretical Neuroscientist Workshop. Very grateful to have the opportunity to host this impressive cohort. We had a glimpse of the future and it looks bright!
1
0
16
When the mouse supremacists tell you that all of primate literature is suspect due to not accounting for uninstructed movements:
In primates, activity in visual cortex is minimally driven by spontaneous body movements. These results confirm the functional specialization of primate visual processing and contrast with findings in mice.
2
0
17
@mortenbc58 @ReichlinMelnick Not just βinvolvement in American cultureβ; immigrants and visitors are an essential component of what America is. Without them, there is no American culture.
1
0
14
Priced out of my Brookline apartment and, to rub salt in the wound, I just found out it's being filled with a bunch of MathWorks devs. #BostonGradSchoolLife.
2
0
14
@ZombieFoodllc @cosmojellyfish Absolutely not but itβs a common mistake to make. Hawaiian is a demonym reserved for those with indigenous ancestry. Itβs a respect thing: Iβm always a βlocalβ but never βHawaiianβ or βnativeβ because I know Iβm on annexed land.
0
0
13
@KaoLeeYang1 Without commenting on how oppression is defined and if itβs a prerequisite for a particular definition of being βunderrepresentedβ by the HHMI, the NIH and US Census consider Hmong (among other Asian ethnicities) to be underrepresented in the life sciences. Link below.
1
5
13
I have literally had a conversation too loudly at the climbing gym about my research and then gotten into an argument with a stranger about whether to use UMAP. I love Boston.
Peak Boston/Cambridge, there are UMAPs and Volcano plots in the train ads ππ€
0
1
15
@andpru Relatedly, I tell undergrads interested in neuro to pick a single major in math/applied math/data science and volunteer in a neuro lab 20 hours a week.
3
0
14
@ebblackwood @drdevangm Iβve always understood this in the context of being inclusionary towards techs and undergrads.
3
1
13
thus it shouldn't be "never use UMAP" but instead "if your latent dimensionality is far from the projected one, be careful interpreting it". This *doesn't* mean UMAP isn't useful for transcriptomics, only that the projected space can't serve as a map in the most literal sense.
1
0
14
I always find freely-behaving macaque studies so impressive and this is no exception! Now what I really want to know is how hard was it to train them to wear mobile eye trackers.
After 2 rounds of applying to graduate programs and 6 years of experimental and computational effort, I am incredibly grateful to present my dissertation work and first, first-author paper on social learning in freely moving macaques, out today in Nature!.
1
0
13
I loved my time βtime offβ but yes, Iβve started to see admits with first-authorships whereas speaking to friends who did PhDs a decade ago, any authorship was rare. Anon grad program this year (not mine) got a thousand apps for what might be 20 offers. Things getting out of hand.
@martinowk Now all of a sudden it is getting harder and harder to get into grad programs without first author pubs, plural. and "experience". postbacc and postbacc-like tech positions are moving from rare to advantageous and are heading towards obligatory.
1
2
13
Growing up I definitely took Hawaii for granted. This place was always just βthat quiet park down the streetβ.
1
0
12
It's always a struggle to embed into 2D but UMAP/t-SNE/LE all do a pretty great job (it's clearest in UMAP) identifying the latent dim! I bumped up the Zipf dim. to make it easier visualize but all methods do very well. Can you guess what number I chose?
How about this set if 10βΆ x 10-dimensional points sampled i.i.d. from a Zipf distribution? Does the manual say anything about that?. Keep increasing the number of points, and watch the medusa get more interesting. Friends don't let friends interpret UMAP plots.
3
1
14
Contributing an intuitive guide to the steps of UMAP I made for our lab's paper. Follow the blue squares up until B vi.
0
0
14