Smerity

22 ms·

121.
Fast and easy baseline text categorization with Vowpal Wabbit(nlpers.blogspot.com)
5 points by Smerity 10y ago|0 comments
122.
It's ML, not magic: machine learning can be prejudiced(smerity.com)
2 points by Smerity 10y ago|0 comments
123.
by Smerity 10y ago
Just to mirror what was said on the thread a month ago when the paper came out[1], if you're interested in FastText I'd strongly recommend checking out Vowpal Wabbit[2] and BIDMach[3]. My main issue is that the FastText paper [7]
124.
It's ML, not magic: machine learning can be prejudiced(smerity.com)
2 points by Smerity 10y ago|0 comments
125.
by Smerity 10y ago
I'm sure many will read into the similarity between the bandwagon for information theory described here and the bandwagon for machine learning / artificial intelligence. While it may be obvious for some, let me decode (one of the
126.
by Smerity 10y ago
Have you seen the (now quite ancient) animated evolution of default Facebook privacy settings[1]? It's insane. It's for this reason that, by default, I assume essentially everything I do online is public. My username is my name, a
127.
by Smerity 10y ago
It's a surprise that Tinder launched Tinder Social just now in the US given that's the main source of the leaked data. Tinder Social was (and remains) opt-out in Australia while he was writing the article. Even if Tinder Social is
128.
by Smerity 10y ago
Some people are noting that TensorFlow evens the playing field, allowing other companies to poach Google's employees. Whilst that is true to some light degree, there are amazingly large caveats - Google still have the major advantage w
129.
by Smerity 10y ago
> Anyways, in most cases, this sort of thing just scales well. It wont do a great job in some cases, but it can do well in many, and if you're hiring tens or hundreds of thousands, thats good enough. That is a hugely concerning comm
130.
by Smerity 10y ago
The European Union are likely to accelerate much of this type of research. They recently introduced regulations specifically targeting algorithmic decision-making and a "right to explanation" for automated decisions. There's
131.
by Smerity 10y ago
The issue is not whether justice can be codified. Let us assume that you can and it's outlined on page 3 of your document. Providing examples to an ML model don't guarantee it will codify anything similar to that. This is a fundam
132.
by Smerity 10y ago
Algorithms _can_ be prejudiced. See [1]. Not being defined with bias or prejudice does not make them impervious to it. If a dataset is biased, as you have stated by "not trusting judges" elsewhere, then the algorithm will follow t
133.
by Smerity 10y ago
Stating that method A is problematic does not automatically mean method B is better. > The reality is humans are far worse Citation needed - especially when comparing against a specific instantiation of a machine learning model. Papers p
134.
by Smerity 10y ago
From "Developing a Practical Forecasting Screener for Domestic Violence Incidents": "We had data from the screening instrument for 671 households. ... For the 516 households with complete data, there was at least one return c
135.
by Smerity 10y ago
As someone who does machine learning, this absolutely terrifies me. The "capstone project" of determining someone's probability of committing a crime by their 18th birthday is beyond ridiculous. Either the author of the artic
136.
by Smerity 10y ago
When I started programming, the basic digital world already existed for me. I never had the experience of typing in a program from a magazine. This is one of those odd experiences where you remember that self-hosting[1] can involve far more
137.
by Smerity 10y ago
I intended these questions as broad examples though some are specific to a given AI system or architecture. You're quite right in that it depends on the task and the model. I was thinking specifically about neural machine translation (
138.
by Smerity 10y ago
The OpenAI team would be best to comment but I've not seen them note any specific intent[1] to research topics related to fairness, accountability and transparency in ML. I use those words specifically as FAT ML 2015 was part of ICML 2
139.
by Smerity 10y ago
Great additions. I can't believe I missed "what algorithm(s) are you using" - it didn't even occur to me people could skip those yet still be taken seriously though that was (obviously) the case with the company I referr
140.
by Smerity 10y ago
That's a really good question and why the initial promise of deep learning was best shown in vision. It is only recently that deep learning in text has given enough performance gains to justify the (large) computational overhead. For m
141.
by Smerity 10y ago
For anyone who is interested in efficiently classifying text, I can't recommend Vowpal Wabbit[1] (VW) enough. It's blazingly fast and has been used in both production and research. It also has a billion options out of the box for
142.
It's ML, not magic: simple questions you should ask that help reduce AI hype(smerity.com)
1 points by Smerity 10y ago|0 comments
143.
by Smerity 10y ago
MetaMind, a Salesforce company | Palo Alto | Full-time | H1B OK MetaMind are a machine learning company. We're interested in machine learning engineers and machine learning researchers who want to push the state of the art in machine l
144.
by Smerity 10y ago
My specific comment re: students was in reference to the perpetuation of "people who were searching 'what is the EU' have already voted on Brexit". That falsely assumed that those searches were from people who had alread
145.
by Smerity 10y ago
I'm glad this was written and has more historical examples than just Brexit. I could imagine a good portion of the questions could be originating from students. If a large political or economic event happens, it's not unlikely for
146.
In deep learning, architecture engineering is the new feature engineering(smerity.com)
2 points by Smerity 10y ago|1 comments
147.
Deep Reinforcement Learning: Pong from Pixels(karpathy.github.io)
189 points by Smerity 10y ago|13 comments
148.
Stanford CS231n deep learning for vision videos taken down due to legal concerns(twitter.com)
8 points by Smerity 10y ago|2 comments
149.
by Smerity 10y ago
I also recommend reading @fchollet's guide on integrating Keras and TensorFlow, especially for those wanting to implement novel components at a lower level :) http://blog.keras.io/keras-as-a-simplified-interface-to-tens
150.
by Smerity 10y ago
This is a pleasant surprise. The more people that work on TensorFlow the better, especially as the DeepMind team will be more aligned with extending TensorFlow's research potential. I am curious how well TensorFlow fits for many of Dee

More ›