physicsyogi

13 ms·

61.
by physicsyogi 8y ago
The Swift for Tensorflow team has added some python interop to Swift. So you’ll be able to, for instance, do an “import numpy” in your Swift code.
62.
by physicsyogi 8y ago
The article says the accuracy of the stethoscope’s predictions is 87%. I wonder what the precision and recall are. Precision saves money, recall saves lives.
63.
by physicsyogi 8y ago
Vancouver
64.
by physicsyogi 8y ago
Same here, except it was my uncle. He worked at the plant in Nevada. I remember speaking to him a few days before the firings and he said Elon Musk was coming to the factory and that was not a good sign. Next thing he knew lots of people we
65.
by physicsyogi 8y ago
I’ve found that some CNNs consistently beat fasttext in terms of model quality. But I’ve beaten those CNNs and fasttext by doing transfer learning with ULMFit and fasts I. But if we’re talking training speed, fasttext is indeed aptly named.
66.
by physicsyogi 8y ago
> You can't sideload an app on an iOS device, you can't run unsigned apps, and there is no version of iOS you can run without the App Store. Unless you jailbreak. This is not quite true. If you write your own app you can instal
67.
by physicsyogi 8y ago
I think GP's instructions work only on Face ID devices.
68.
by physicsyogi 8y ago
There's a sci-fi novel called _The Truth Machine_ that looks at what happens in a society when a perfect lie detector is introduced.
69.
by physicsyogi 8y ago
I used Lua's C API a couple of years ago to write some lightweight JNI code to use Torch from Java. The API documentation was so clear that it was relatively easy to write. Since the API is stack-based the project ended up being a good
70.
by physicsyogi 8y ago
The paper's abstract suggests it was not: https://jamanetwork.com/journals/jamainternalmedicine/fullar... >> Design, Setting, and Participants In this population-based prospective cohort study among Fr
71.
by physicsyogi 8y ago
P-Diddy sold drugs for literally two days. http://hiphopdx.com/news/id.10271/title.diddy-opens-up-about...
72.
by physicsyogi 8y ago
This phone sounds interesting. It also reminds me of the Pied Piper platform from the Silicon Valley show.
73.
by physicsyogi 8y ago
At the bottom of their homepage there's a link to an online APL interpreter: https://tryapl.org/
74.
by physicsyogi 8y ago
The only example of a Clojure team migrating to Haskell is the Duckling team, a little while after Facebook acquired them. Among other things, they wanted a language with better support within Facebook. https://medium.com/wi
75.
by physicsyogi 8y ago
The penthouse in question reminds me of how houses are staged. It looks sterile and as the article says, "free from the clutter of everyday life".
76.
by physicsyogi 8y ago
"Centuries" is a bit of a stretch, right? Or am I misunderstanding your comment? Some of the most widely used algorithms in machine learning, for training or prediction, are much younger than that: - backpropagation (1970s) - Vite
77.
by physicsyogi 8y ago
There's also the isotope Helium-3, which is far more scarce than Helium-4 and is rapidly running out. It's used to detect radiation and to reach ultralow temperatures in low-temperature physics research and medical cryogenics. It&
78.
by physicsyogi 8y ago
Wharton is at UPenn. Penn State's business school is Smeal.
79.
by physicsyogi 8y ago
Cryptonomicon and Snow Crash were great. Others you might: - The Story of Your Life by Ted Chiang (superb short story that Arrival was based on) - The Three Body-Problem (Cixin Liu) - Dune (Frank Herbert) - The Parable of the Sower (Octavia
80.
by physicsyogi 8y ago
I generally prefer to see neural networks depicted so that the input is on the left and the output is on the right. This is probably because I think of time on an axis as flowing left to right.
81.
by physicsyogi 8y ago
They're calling it "text neck". https://www.washingtonpost.com/news/morning-mix/wp/2014/11/2...
82.
by physicsyogi 9y ago
Enriching Word Vectors with Subword Information: https://arxiv.org/abs/1607.04606
83.
by physicsyogi 9y ago
> The goal of this project is to build a functional database. > neon is a persistent set storing quads ie. 4-tuples made of: > (graph, subject, predicate, object) Sounds interesting. And reminds me of Datomic. (Edit: formatting)
84.
by physicsyogi 9y ago
Clojure has Flare[1] and Cortex[2] for machine learning. Clojure is also a great language for data processing. Python, of course, gives you access to a host of useful libraries for NLP (SpaCy[3], NLTK, Textacy), machine learning (Scikit-lea
85.
by physicsyogi 9y ago
Maybe look into ConceptNet Numberbatch: https://github.com/commonsense/conceptnet-numberbatch/blob/m...
86.
by physicsyogi 9y ago
There's a nice Jupyter kernel for Clojure: https://github.com/didiercrunch/lein-jupyter .
87.
by physicsyogi 9y ago
This is like Jupyter Notebooks for Mathematica.
88.
by physicsyogi 9y ago
Microsoft used stack ranking for some time and from what I've read it caused a lot of problems and was considered to be negative overall. It led to a cutthroat environment where employees were competing with each other (and not focusin
89.
by physicsyogi 9y ago
>This is a problem mostly because of the way we’ve organized the web. There are many clients that want to get content and use programs and only a relatively few servers that have those programs and content. When someone posts a funny pic
90.
by physicsyogi 9y ago
I didn't know that Grapher was still around.

More ›