kanak

> MIA Meeting: Low dimensional embeddings of words and documents (and how they might apply to single-cell data)

Leland McInnes Tutte Institute for Mathematics and Computing

Over the last decade the field of Natural Language Processing (NLP) has been overtaken by neural networks and deep learning. The latest models and algorithms, from word2vec to Google’s Universal Sentence Encoder and BERT, can perform seemingly magical feats and provide powerful tools to understand and analyze text documents. This talk will seek to pick apart word and document embedding techniques from NLP, removing the neural networks and instead working with linear algebra and dimension reduction on large sparse matrices of counts that is very similar to a large amount of single-cell data. Such an approach can achieve results on par with neural network approaches, but are both simpler to understand the inner workings of, and are generalizable to much more diverse domains than NLP, including, hopefully, single-cell research. In particular all we really need to apply these kinds of techniques are large sparse matrices of counts and some notion of locality or cooccurrence of the feature columns of that matrix. The goal of this talk is to hopefully open up some of the ideas that have been so effective in NLP and make them usable and accessible to a wider audience working with more diverse types of data, and initiate a conversation about what is required to make such techniques useful for the single-cell community.

A recent paper trying to systematically apply the idea to databases "Towards Systematic Index Dynamization" ( https://bpb-us-e1.wpmucdn.com/sites.psu.edu/dist/b/123163/fi... )

https://github.com/apache/lucene/blob/84cae4f27cfd3feb3bb42d...

Kubernetes Up and Running by Brendan Burns, Joe Beda and Kelsey Hightower (published by O'Reilly).

Allegro Prolog [1] is based on Norvig's implementation, although I'm sure they've done a fair bit of optimizing.

[1]: http://www.franz.com/products/prolog/

Are you talking about extensions that I already have installed or any extension on AMO that might not work with the new version?

http://twitter.com/#!/csoghoian

There's also "design concepts in programming languages".

https://userscripts.org/scripts/show/97052

I don't know about the junior lab (i'm in eecs), but when I take classes that are difficult (which has been pretty much every semester for the past 2 years :(), I am working overtime on these classes. Besides attending lectures/recitations during the day, I spend every evening, and all of my weekend coding for the projects. Having projects and labs that took over 20-30 hours to complete were the norm, and in a particularly bad semester, I'd have one of those due almost every two weeks.

Initially, there would be people who could get by without putting in as much effort because they already knew the material, but as you go higher and start talking about lab classes (where the work is punishing regardless of your knowledge/experience) or graduate classes (very intensive, almost no one can coast), you don't see people coasting.

First off, this is for students in Aero-Astro program. It's not something that all engineers take; it's simply the starting courses for the aero astro people.

Second, even though the name suggests that it's some course that tries to teach you the intersection of all the topics, it's actually teaching you the union of the topics. From the course overview:

"Over the course of each semester, students engage in seven disciplines of study, listed below. Each discipline is taught for a fraction of the semester, through a series of lectures. When one discipline concludes, students are quizzed and begin a new discipline."

Similarly, if you look at the calendar ( http://ocw.mit.edu/courses/aeronautics-and-astronautics/16-0... ). You see that the students are essentially taking each sub-course separately.

It's more for scheduling convenience and to have the students up and running with all fields in Aero-Astro than to do some grand unification of these topics.

I've been using org for about 2 years to take notes... I've taken several hundred pages worth of notes using it by now. I've been using it for about a month as my task / appointment manager, and I'm impressed by its flexibility.

I was in the same spot when I started because there's a lot of things that Org is capable of. I used the following resources when I was learning Org. You might find them helpful:

1. Carsten Dominick (lead dev of Org-Mode) talks about Org at Google: http://orgmode.org/talks/index.html#sec-1

2. "Power users describe their setup" section on the worg tutorial list: http://orgmode.org/worg/org-tutorials/index.html

By this point, you'll have atleast some idea about how org can be used. If you decide to delve in further:

3. Sacha Chua's blog posts on Org-Mode: http://sachachua.com/blog/

4. Org compact guide: http://orgmode.org/guide/index.html

This is definitely true. For example, if you put a video show you got off of eztv in your dropbox, the upload is instantenously complete. Why? because someone else has already put that file into dropbox so you don't need to upload it as well.

In comparison, amazon is offering 100 GB for the same $100 a year, and even has cheaper paid plans (e.g. $20 per year for 20 gb).

So, dropbox is currently expensive on two fronts; Amazon gives more GB/$ and also has lower-priced plans for more modest needs.

This is a big win for some users, like kunjaan and me, because we haven't been too thrilled with Dropbox's syncing abilities (tons of "conflicted copies" everywhere). Furthermore, a large portion of our data doesn't need to be synced; it simply needs to be available. This includes data like ebooks, music collection, pictures.

It's a difficult journey, but it is a scenic tour of a lot of the core concepts in computer science.

[1] http://eli.thegreenplace.net/2008/04/18/sicp-conclusion/