RyEgswuCsn

I have been using Docker powered desktop environment for development on WSL2. Didn't have any issue except that I couldn't access the containers via their internal IPs inside WSL2.

Sometimes I feel that the free market argument doesn't apply to media platforms. This is because the effectiveness of a media platform (or any mass media, for that matter) is largely determined by its market/usage share; in other words, smaller or alternative media platforms cannot be considered as a working substitution to the dominate platforms. And since the user distribution of social media tends to follow the power law (i.e. the "can't switch 'cause all my friends are using it and they won't switch due to the same reason" phenomenon), it is almost inevitable that a single platform will eventually monopolise the sector if left unregulated. banning a user from using the most popular media channel would mean he/she is no longer able to communicate his/her opinion effectively even he/her is able to choose a less popular alternative.

tl;dr: one can always switch their phone if they don't like Apple because the usefulness of a phone doesn't depend on its popularity; one can't practically switch their social media platform if they are banned because the usefulness of the social media depends on its popularity, and the most popular platform typically dominates due to the power law.

Some of your points feel off to me:

    "they insist Chinese law is used outside China, and lecture people how they should refer to China in public;"

Not really sure what examples you had in mind, but I would imagine most communities would have a preferred way for other people to address them; I don't think it's wrong for them to demand that. If you don't like China's demand then you can refer to it however you want; but they also have the freedom to dislike those who do that.

    "they operate concentration camps;"

In my unpopular opinion: I would call them brainwashing camps, or re-education camps if there is no need to appeal to Western audiences. I don't think I have seen evidences that people have been systematically terminated in those camps, which I believe is a key characteristic for concentration camps.

    "the Chinese language has no word for *mutuality*"

Did you mean what you said seriously/literally? Or is it a phishing question to see which commenter speaks Chinese? ;)

    "China benefits from free speech"

I would argue that they didn't benefit much from "free speech", otherwise China would probably be mentioned in a more positive light nowadays. IMHO, China's state machinery is pretty good at internal censorship and propaganda due to their long authoritarian history, but their methods for manipulating/spinning narratives is rather unsophisticated compared to societies with more democratic traditions.

    "Winnie the Pooh is extremely thin-skinned and can have games like Devotion removed from Steam and GOG even after the reference is removed from the secret area in the game;"
    "there are many instances where businesses with Chinese shareholders were pressured not to publish anything mildly critical of China."

Capitalism has no backbone, I guess?

    "BBC team visiting China investigating covid19 faced obstacles reminiscent of The Truman Show;"

Your neighbour doesn't like you, and he thinks you are guilty of theft. You told him you have not done that and turned down his demand to enter your home to investigate. Your neighbour informed the police, and they come to your door with a warrant; this time you have to let them in. Is BBC the world police? If so then I guess China is at fault here.

    "WHO scientists investigating covid19 recently not permitted into China;"

I think WHO is much better qualified than BBC for investigating this, so not a great move from China. They should let them investigate.

    "China doesn't even formally frown at corruption;"

I can see you don't know China at all. China does formally frown at corruption. The problem is China informally doesn't frown at corruption.

If Wikipedia is to be trusted [1]:

"AIDS was first clinically reported on June 5, 1981, with five cases in the United States"

"Both HIV-1 and HIV-2 are believed to have originated in non-human primates in West-central Africa and were transferred to humans in the early 20th century."

If HIV, a relatively less infectious virus, can make its way across the Atlantic Ocean to land on America shores, I wouldn't be surprised SARS-CoV-2 could do the same. Another explanation could be that SARS-CoV-2 is a lot more common than we think --- many people might have caught it before but it didn't manage to spread, but this time the virus just hit the jackpot.

[1]: https://en.wikipedia.org/wiki/HIV/AIDS#History

Maybe they do secretly handle coronaviruses in the provincial administrative office; they can't prove they don't anyways.

There are thousands of wet markets in China, but very few provincial capital cities (Wuhan is one). Out of all the wet markets in China, the virus just happened to originate at the one next to the provincial administrative office? That alone is enough evidence to make the provincial administrative office the most likely origin of the virus, right?

Would this work?

  return Optional.ofNullable(nullableVariable).map(v -> v.someMethodCall()).orElse(false) ? true : fallbackIfNullMethodCall();

Very nicely written post. I particularly like how you attached a link to your codebase on repl.it so anyone who is interested can tinker with the code.

One thing I have been wondering for some time is whether the vanilla RNN can learn negations (i.e. 'not good' == 'bad') and valence shifts (e.g. modifier words like 'very' --- they do not carry sentiment connotations themselves, but may amplify/dampen the sentiment of the words they modify; negations like 'not' can be considered as a special-case valence shifter where it inverts the sentiment of the following word).

My suspicion is that vanilla RNNs are not capable of modelling negations and valence shifters since they make inference on the sentiment of a sentence by 'adding up' the sentiment connotations of its constituent words --- negations and valence shifts, however, works more like multiplications than additions.

I see you already have such examples in your dataset so I thought I'd do some experiments. I simplified your original dataset to the following:

  train_data = {
    'good': True,
    'bad': False,
    'not good': False,
    'not bad': True,
    'very good': True,
    'very bad': False,
    'not very good': False,
    'not very bad': True
  }
  
  test_data = {
    'very not bad': True,
    'very not good': False
  }

While the test cases do not reflect how people actually speak, the hope is that the model should be able to apply its learning to infer their sentiment. For me, however, it would seem the training failed to converge with the default parameter settings (hidden_size=64).

It would be interesting to see how other RNN architectures (e.g. LSTM, Transformers) fare with negations and valence shifters.

P.S.: When calculating softmax, it is better to use the built-in functions or at least do the log-sum-exp trick to prevent under-flowing.

I find the title of the article rather exaggerating...

As of the first difference pointed out in the article, one of the CS224D lectures on word2vec did addressed it:

https://youtu.be/aRqn8t1hLxs?t=2650

It was also mentioned later in the lecture that having two vectors representing each word is meant to make the optimisation easier (so it's kind of a trick); at the end, the two vectors learnt will have to be averaged over in order to reach a single vector for each word.

To be fair, the fact that each word is represented by two vectors was also mentioned in the original paper describing word2vec:

https://arxiv.org/pdf/1310.4546.pdf

On page 3, just beneath equation (2).

Why so surprised?

If you are using Python with Jupyter notebooks, another way to create graphs with ggplot2 is to interface directly with an R session using the rpy2 package. A nice tutorial can be found here: https://www.linkedin.com/pulse/interfacing-r-from-python-3-j...

Yes they did. I noted them in another comment.

Agreed. Also noted three potential flows:

1. They applied pandas.qcut over the entire returns (i.e. training and validation set) when generating the target performance labels, which could compromise the validation.

2. The ordering of news is actually important when it comes to forecasting the movements of the market; their training/validation split was done after shuffling the entire dataset, which means the model will have access to information that shouldn't be available to it during training.

3. It makes more sense to use a training-validation-testing setting rather than a training-validation split when reporting the model performance in order to avoid inflated results due to all the hyper-parameter tuning.