jointpdf

Alternatively, carbon tariffs: https://www.niskanencenter.org/border-adjusting-imported-pro...

Still, from the point of view of the consumer it’s “direct” (in ordinary language, not economese). As in: “concretely and transparently makes product X more expensive for me”.

With an upstream carbon tax, it’s unclear how it affects the prices of consumer goods since it has to filter through the entire supply chain first. It’s not like “this pack of gum produced 3.14 pounds of C02, so you—dear consumer—will be taxed an additional $0.42”. That would be a nightmare to administer.

Example: if you crudely crunch MNIST from a [60000 x 784] matrix down to [60000 x 15] matrix (15 is arbitrary here—anything smaller than the original 784 works) using random projection, the JL lemma tells you how badly you mangled the similarity structure of your dataset in the worst case scenario. It turns out it’s not so bad, so we can use DR to compress our data matrix and then confidently do useful things with the vectors it contains (like search for similar ones).

ScaNN is an approximate nearest neighbor (ANN) search algorithm. There is an entire zoo of them, but they all have the same goal: given a query vector q and a data matrix X (where q has the same number of elements as a row vector in X), find the k vectors in X that are “most similar” to q, without exhaustively checking all the possible matches. Most similar could be any similarity or distance measure, but in this case it’s the simple inner product.

To do this more quickly, the algorithm uses a type of vector quantization (i.e. sort the high-D vectors into buckets), which is similar to k-means. K-means can also be construed as a dimensionality reduction technique, you are just reducing to a single dimension. That way you’re only looking in buckets where there’s hope of finding close matches.

This works great in real life with socks—if you’re trying to find a match for a stray black dress sock, then 1) make sure you presort your socks so that they are approximately binned by similarity and 2) look in the bin that contains socks most similar to your “query sock” (i.e. don’t look in the colorful athletic sock or plain white dad-sock bins). I’m not actually kidding, if you have a billion unmatched socks in a box, try this and you’ll see.

Anyways...as for this particular algorithm, someone else may be able to explain the details better. But my coarse understanding is that the loss function (which measures how good the bucketing is—how purely you pre-binned your socks) was defined in such a way that accounts for the density of the region in question—a neighbor in NYC means something different than a neighbor in Wyoming, so you need more granular buckets for denser regions. However, I thought this was essentially how almost every ANN algorithm worked, so how this one blows all the others out of the water (faster for a given recall value) is mystifying to me.

[PS—someone please correct me if any aspect of my loosey goosey explanation is inaccurate]

A carbon tax is overwhelmingly considered to be the most efficient solution by economists for Econ101 reasons—pricing negative externalities (correcting a market failure is more efficient than doing nothing), and enabling competition and comparative advantage (more efficient than command-and-control policies).

The idea of taxing things we want less of is known as a Pigouvian tax, by the way—and it’s considered a no-brainer even by the most conservative economists (e.g. Mankiw).

> good luck getting people to swallow a doubling of their tax bill even if you give it back.

Not so! Public opinion on climate change has shifted dramatically in the last several years. This poll from last year (and from a top Republican pollster to boot) found that people support a carbon-tax-and-dividend policy 4-to-1. 75% of GOP voters (!) under age 40 support this type of policy. Read all of the results, they are remarkable. It’s a winner in the eyes of both economists and the public. https://www.clcouncil.org/media/Luntz-Carbon-Dividends-Polli...

I think it’s better to just assume nothing unless you have direct knowledge that the employee is doing something blatantly unethical like working at another job (has happened before).

I was taught this as a way to control some very acute anxiety from hypochondria that took over my life for ~6 months before I could get treatment. Apparently it worked, since I haven’t suffered from that type of anxiety since.

Sometimes a rant on a topic brews in my head for weeks or months, and I will uncork it on a random passerby that brings up the subject—which happened to be you this time.

But, I’ve had coworkers who like clockwork sneer at anything a data scientist wrote. “Why did you do it that way?”. When asked for advice on how to improve it, they huffily say nevermind. It’s ingratiating as hell.

I’ve gone out of my way over the years to make learning data science skills as approachable as possible for uninitiated (giving trainings, providing customized learning paths based on someone’s background, offering encouragement), and yet this is almost never reciprocated by engineer types. It’s always just, “data scientists can’t write production quality code”, with no explanation of what production quality entail, or without consideration of the fact that notebook-based data science can have advantages over perfectly modularized code with a battery of tests. See the comment above: “I'm not even sure what to recommend for developing good software judgment and habits.“. It’s like a chess coach admonishing their subject to simply “think harder”. Not helpful.

When curious and open-minded data scientists and software engineers work together, it can be magic. When people snipe at others for their “shitty” skills, it creates a petty and toxic environment.

This comment comes off as a bit of an admonition, but I would greatly appreciate a list like TFA for data scientists looking to shore up their fundamental CS and software development skills.

(PS — The first book I read when teaching myself R was R Inferno, so that ain’t it.)

I tend to not be a cover-to-cover reader, so I usually deep dive into a single topic for a while (e.g. forecasting, information retrieval) and read papers/tutorials/chapters related to that topic and the math concepts related to it.

PS I feel like impostor syndrome is so common among data scientists because there is so much material that feels like “must know”. Don’t feel like you need to memorize thousands of textbook pages to be effective, and you could spend a lifetime mastering any of these individual subjects. JIT learning is a great skill to have.

* Vectors, Matrices, and Least Squares — IMO the best beginner-friendly and applications-focused intro to (or review of) linear algebra. Covers a ton of fundamental ground while keeping things consistent and concise. Lots of exercises and a Julia supplement book. (http://vmls-book.stanford.edu/)

* Mathematics for Machine Learning — good coverage of the most important math concepts relevant to ML (https://mml-book.github.io/)

* Forecasting: Principles and Practice — best overall resource on forecasting that I know of; R focus. One of a zillion great R/data science books, virtually all of which are open and well-written. (https://otexts.com/fpp2)

* Dive Into Deep Learning — can’t personally vouch for this one but it looks comprehensive; numpy/PyTorch/TF focus (https://d2l.ai/)

* Speech and Language Processing — clear introduction to all things NLP, nice flow (https://web.stanford.edu/~jurafsky/slp3/)

Not long after that I (spoiler alert) needed a psychiatrist. It took me 30 calls to find someone with availability. I punched a hole in my wall around call #20.

As recently as last Fall, I was still spinning my wheels, delaying the unpleasant search process for a therapist. I was so frustrated with “the system” that it short-circuited any desire to make the effort. There were at least four major barriers:

1) communication anxiety — worried about explaining myself and my problems to a prospective therapist. [addressed in my previous comment—keep any introduction email/call extremely simple.]

2) financial anxiety — as with all healthcare in the US, the amount you pay is a random variable. Maybe it will be a waste? Maybe I will go bankrupt? [if you have insurance, find someone on your plan using the Psychology Today search filters—they also often list the cost. if not, there are other free or low-cost resources available, but it’s context/location dependent. some potential options include the VA, school/college therapy, counseling programs @ your employer, therapists in-training, various community/nonprofit organizations. ask a trusted friend to help you search, it can alleviate some of the frustration. you can ask me.]

3) avoidance anxiety — embarrassment around explaining my life story. Why and how should I explain the fact that I smoke weed 10x a day (a true 10x developer over here) to cope with my issues, when I could simply roll another joint? [one role of the therapist is to be a non-judgmental sounding board and co-strategizer—they aren’t your coworker, parent, partner, etc. this is a hard mindset to learn, even for me to this day. I find myself saying things I think my therapist wants to hear. as a corollary: ER docs have seen the full Cartesian product of injuries from {objects} x {orifices}. whatever you come in with is not as unique or embarrassing as you think. they do this all day long.]

4) perfectionism — “Yeah I’m sure this therapist can help someone get over a breakup or their dead cat, but I’m a special brand of fucked up and they won’t get it. I need to find a someone that is perfectly specialized in my personality/problems, or it will be a waste. Also they need to have a PhD from a good school, female, not too young or too old, not ...”. [just a reminder that this is not a search for your soulmate. a student therapist or MSW may be more effective for you than a famous professor with books and accolades (see above). anecdata: the two therapists that I hand-selected didn’t pan out for me at all. the two that were chosen for me at random were a wonderful fit.]

Finally, I would push back on the idea that remote therapy is ineffective. Telehealth in general is proving to be pretty effective, and with therapy they don’t need to interact with any physical structures. The only thing that is slightly diminished is nonverbal cues, but that’s not insurmountable. There are a lot of advantages: significantly more options (no location constraints), easy to schedule and keep appointments, can be more comfortable to not be physically present, and no stressful driving. I recently did a therapy appointment call from the Appalachian Trail!

Once you put on the HeatTech extra warm leggings, you’ll understand why superheroes wear tights—they’re the bomb.

First of all, don’t stress about explaining “what’s wrong” to a therapist before you meet them, or let this become a barrier to you getting started (I fell into this trap for years myself). A simple “Hi my name is X and I’ve been been thinking about finding a therapist to work on some issues with. When would you be available for an initial call?” is perfectly sufficient.

A thing people say about therapy is that it’s like a bit like dating. Don’t be discouraged if it takes a few tries to find someone who is accepting new patients and feels like a good fit (most will offer a free consultation as well to help you evaluate this). Give them a fair shot, a few sessions at least (I was skeptical of a therapist I was randomly assigned ~10yrs ago, but he swiftly cured a very acute case of hypochondria that had bubbled up out of nowhere. However, you should move on to the next one if it really isn’t working—unlike dating, your therapist won’t be hurt (or if they are, then they have the tools to deal with it right?). This is probably the best directory for finding local therapists: https://www.psychologytoday.com/us/therapists

Most will give you a set of questionnaires to complete before or during your first appointment (e.g.: https://www.uwgb.edu/UWGBCMS/media/Continueing-Professional-...). This gives them a coarse baseline of what’s going on.

The conversation simply evolves from there. Over time, you will occasionally pull out surprising memories and insights from the depths of your mind like the proverbial Mary Poppins bag. There’s a lot in all of us to unpack, and they are there to facilitate that process.

A common attitude (again, one I’ve suffered myself) toward therapy is: “What will a therapist tell me that I don’t already know? I can solve these problems on my own if I actually tried.” Maybe, but the operative word phrase here is “actually tried”. Actively reflecting on your thoughts/emotions/actions and vocalizing them to another person for an hour is challenging, but it is therapeutic in and of itself. It’s very hard to do on your own, and sometimes take a probing question or two to dislodge a blind spot.

Also, if you want to listen to some real therapy sessions (intermixed with lecture content on the theory and research behind a common type of therapy), these lectures on Cognitive Behavioral Therapy are excellent. It follows a few different people from the very beginning (skeptical of or unsure about therapy or what their deeper issues are) through their process of improvement via weekly sessions/assignments. It can really help demystify therapy for the unfamiliar: (https://www.audible.com/pd/Cognitive-Behavioral-Therapy-Audi...)

PS—Keep asking good questions.

I went to a private school for 1st+2nd grade (this is Michigan, in the 90s), and we were allowed to roam around unsupervised in the woods during recess. That includes the dead of winter. We built forts out of tree branches, constructed igloos and snow tunnel, etc. Occasionally we did outdoor lessons, like making a to-scale model of the solar system.

I’m not suggesting those activities necessarily, but it seems to me that there is the possibility of moving outdoors for a significant portion of the day. Even in the absence of a lesson plan, kids still get the benefit of social interaction, experiential learning, etc.

https://www.semanticscholar.org/paper/Cricket-chirping-algor...

This is the kind of statement people should put in their dating profiles if they want fewer but better matches. No, I’m not teasing you (ok just a little). Show quirky elements of yourself, just keep the mood sunny rather than salty.

Bonus—An excellent interactive Spacy course from Ines Montani (also includes a template to build similar courses!): https://github.com/ines/spacy-course

- Snorkel (training data curation, weak supervision, heuristic labeling functions, uncertainty sampling, relation extraction): https://github.com/snorkel-team/snorkel

- AllenNLP (many pretrained NLP research models for tasks beyond text classification, model training and serving, visualization/interpretability utilities): https://github.com/allenai/allennlp

- Spacy (tokenization, NER/POS + tagging visualizer, pretrained word vectors, integration with DL models): https://github.com/allenai/allennlp

-huggingface Transformers (latest and greatest pretrained models, e.g. BERT): https://github.com/huggingface/transformers

- ...or a barebones “from scratch” solution in less than an hour with a Colab notebook and scikit-learn (preprocess text into tf-idf vectors, LSA/NMF to generate “document embeddings”, visualize embeddings with t-SNE/UMAP [facilitates weak supervision/active learning], classify with LogReg/RF/SVM/whatever). You could also tack on pretrained gensim/TF/PyTorch models quite easily as a next step. But this basic flow quickly gives you a handle on your corpus.

By the way, the docs for DeepDive (the predecessor of Snorkel) are some amazingly detailed background reading: http://deepdive.stanford.edu/example-spouse

(Read: If X and Y are normally distributed random variables with mean 0 and standard deviation 1, then X/Y is equivalent to a Cauchy(0,1) distributed RV. This is a useful equivalence to know if working with standard normal RVs, which is often. E[X/Y] does not exist!)

If X ~ Cauchy(0,1), then X ~ normalized student-t distribution.

(student-t puts the “t” in t-SNE, where it acts as a weighting function on the Euclidean distance between points. UMAP uses a parameterized version of this weight function that is essentially a generalization of the Cauchy distribution: https://jlmelville.github.io/smallvis/umap.html)

Of these, I have listened to the Breathing Meditation (5 min) and the Body Scan for Sleep (13 min) the most and would recommend them. Maybe keep the 5 minute one open in a tab, and if you find yourself in a procrastinatory tab-switching vortex, break and listen to it.

Here are some longer and more “themed” guided meditations from the meditation guru Tara Brach. She also does weekly livestreams if you’re into that: https://www.tarabrach.com/new-to-meditation/

If you want to learn about the science behind how meditation changes your brain, look at “The Science of Mindfulness” from The Great Courses (it’s on Audible): https://www.amazon.com/The-Science-of-Mindfulness-audiobook/...

Aside from meditation, as a gold medalist procrastinator and long-time (~20yrs) Adderall user, I’d caution you against seeking that out to help you with the problem you describe. It can send context-switching behaviors into overdrive and cause a host of other problems. That said, there’s no shame in taking them if they’re right for you (definitely would encourage you to find a doctor that’s comfortable treating ADD etc, possibly a specialist).

For managing tasks, this the only method that’s effective for me: get some {red, yellow, green} post-it notes (color = priority) and a sharpie. Write down every task that is weighing down your mind (this is the basis of the GTD method). One task per post-it, and write down the estimated time in minutes to complete. If it’s a long task (refactor script.py), I convert it into X number of 30 time blocks that I want to spend on it, not necessarily expecting completion by the end of that time. Use the pomodoro method to knock out tasks, and draw a big ass check mark on the post-it when you finish one (feels great!). I also organize the tasks on a whiteboard with the following rows as categories: {health, life, work, finance, education, projects/other}.

This is my procrastination eradication method. There are many like it, but this is mine.

> “Others pointed to the dualism between black and white as representing evil and good, concepts that have been around since the dawn of civilizations, long before racial divides even existed between humans.”

The supporters of the current terminology use this to bolster their argument, but I believe it’s also at the core of the argument against the use of terms like blacklist/whitelist. The idea is that the black vs. white = evil vs. good (like, MTG card colors) can “spill over” into dark skin vs. light skin = evil vs. good. This is an interesting article related to this effect: https://www.scientificamerican.com/article/the-bad-is-black-...

Ok fine, I’ll take a weakly-held stance: this seems like a bit of a stretch, and distracts from the more serious ways in which systemic racism manifests itself. On the other hand, I see no need to become outraged—words fall out of favor all the time (go play with the Google n-grams viewer to see for yourself).

/s