v1n337

The Youtube channel 'Worth It UK' reviewed this recently.

https://www.youtube.com/watch?v=cFvL4HLV1vw

Yes, quite correct! The vectors are dense representations.

But the features aren't individually interpretable, in practice. For instance, the 'gender' of a word may have it's signal scattered over several features/dimensions of the learned vector.

I love this blog!

This is less of a dev blog, though, and more of a research paper blog.

> The fourth actor is the media. The coverage of the memo has been atrocious.

The irony.

Nice post. Thanks for sharing.

I have to point out though, that it's a bit dangerous to measure classifier accuracy as the percentage of correctly classified samples when you've no idea how the test data is skewed in favor of one class vs. the other (for binary classification, and can be generalized to multi-class problems too).

It's always much better to represent accuracy as the F1 score[1] or to just examine a confusion matrix of the predictions[2].

[1] https://en.wikipedia.org/wiki/F1_score [2] https://en.wikipedia.org/wiki/Confusion_matrix

The GloVe project just happens to have pre-trained models on their official page.

Word2vec do have a Google News corpus model on their official page, but there are many more trained word2vec models in the literature.

You're right. It's a shallow single network with the weight embeddings of words at the intermediate layer extracted as the vector-space representation of a word, depending on its context.

Am I missing something here?

None of the examples they've cited at the bottom of that post are actual contributions to a popular open-source project, but seems more like stuff that's been built as a standalone project.

> asked a fake question

What does this even mean?

> The activist does it all the time.

Do you have a source for that activist's history of RTI requests?

> Also, Modi never claimed he was registered to sell tea.

So, he claims to be an unregistered tea-seller? Because that would make him a tax evader. (http://www.hindustantimes.com/india/modi-talks-development-r...)

> Had the Congress government been so proactive in 55 years by mandating registration of tea sellers, India would be much safer, more cleaner, and less corrupt.

The Congress is incompetent, no doubt. But does that absolve Modi of tax-evasion now?

> Fun fact : That activist is from Congress party.

Sure. What difference does that make to the outcome of the RTI request?

> Another fun fact : Such activists do not disclose the exact wordings of the RTI query, they intentionally post ambiguous queries and use the reply to create sensational headline. I can challenge any of the Modi haters to share their exact RTI queries before they make sensational headline.

The article says "A Congress supporter and social activist Tehseen Poonawalla, had sought information under the Right to Information (RTI) Act from the railway board about whether there was any record, registration number or official pass issued to Modi allowing or entitling him to sell tea on trains and at stations." I don't really see how the wording here could be ambiguous, but I'd be happy to read an explanation.

I know there are plenty of political shills on online forums, but I think it's our collective responsibility to hold political leaders to high standards of virtue, like truth, something Narendra Modi has been fairly inconsistent with.

http://www.thehindu.com/elections/loksabha2014/modi-declares...

https://thewire.in/32800/controversy-over-modis-educational-...

http://zeenews.india.com/news/nation/no-climate-change-says-...

Factually incorrect.

This is not the first time that a non-Congress party has been elected to power in India. The NDA government was in power during 1998–2004, a period including the Kargil war. (https://en.wikipedia.org/wiki/Bharatiya_Janata_Party#NDA_gov...)

Modi was not a chai-wallah. His father owned an unregistered tea-shop at a railway station. (http://timesofindia.indiatimes.com/india/No-official-record-...)

Isn't VSCode Electron based too? After reading this article[1], I've shunned desktop apps using Electron, because it just seems wasteful in terms of memory. I've switched back to Sublime.

[1] https://josephg.com/blog/electron-is-flash-for-the-desktop/

Not, OP, but I've worked for an ad-agency in a software-dev role.

Their primary code base is written in Java, monitoring scripts in Python/Ruby, and tracking pixel code in JavaScript, so I would have days in which I coded in all 3, albeit for separate use-cases.