Hacker News Reader

lapinkon Netflix's UX design is keeping people up at night

Can’t people be grown ups and stop blaming UXs for their behaviors?

lapinkon Comparison of state-of-the-art music source separation models on Californication

Speech source separation has gone a long way, thanks to Yi Luo amazing work. With Dual Path RNN, he now achieves almost 20 Signal to Noise Ratio for 2 speaker separation, see [1]. This is a bit of an artificial setting though, only two speakers and they are manually mixed together. I'm not sure if there is any good dataset of speech source separation in real environments (an airport, restaurant etc).

[1]: https://arxiv.org/pdf/1910.06379.pdf

lapinkon Comparison of state-of-the-art music source separation models on Californication

Author here, this is part of the release of Demucs, you can find more information on my repo: https://github.com/facebookresearch/demucs

lapinkon A Study of Meditation Under the Influence of Psilocybin

Everything is impermanent. Through your practice, develop perfect equanimity. Anichoooaaaa

lapinkon The idea behind dicts being ordered since python3.6

The mail is from 2010 but it was only implemented in 2016, I wonder why. Found the link in https://github.com/python/cpython/blob/master/Objects/dictob....

lapinkon No need to cut down red and processed meat, study says

Please still consider reducing your meat intake for the planet: https://www.sciencedirect.com/science/article/pii/S092181811... If you want to help save Amazonia, this is definitely the easiest course of action.

lapinkon Deep Learning for Symbolic Mathematics

Indeed...

lapinkon Deep Learning for Symbolic Mathematics

There exists many such bases, you could also take all the monomes and approximate a function by its Taylor development. However, it does not mean that such bases can be efficiently approximated in a reasonable dimension, nor that conversion from/to textual representation is easy. A deep learning network would achieve those points.

lapinkon Deep Learning for Symbolic Mathematics

It is a possibility that there is a natural vector space embedding for functions in which integration/derivation is a simple operation. A deep learning network could find such an embedding.

lapinkon The beauty of functional languages in deep learning – Clojure and Haskell

Thread level parallelism is different from GPU parallelism. Different threads can perform completely independent operations at any time. GPU threads must do exactly the same operations, but on different memory locations, at all time. In exchange for this rigidity, we can pack a lot more of them on silicon than CPU. A CPU thread is like a complete individual that can do anything they want. A GPU thread always is part of a pack, and they all move together.

The nice parallelism allowed by Clojure is for CPU threads not GPU threads. It would still need to rely on an external library for tensor operations, for instance ATen [1], the C++ backend of PyTorch.

On the other hand, Functional Programming can be useful to describe the model at a higher level and better handle the scheduling of each component (Convolution, LSTM, etc) on GPU. When training model, the batch size already allows near optimal usage of a GPU cores, however when doing evaluation, this becomes more relevant.

[1] https://pytorch.org/cppdocs/

lapinkon Inbox is about to die, and Google still hasn't brought its best feature to Gmail

When asked in an Inbox user survey what I liked the least about inbox, I answered “ that it is closing down :’( ” I’m still hoping for them to change their mind and I will be using it until the very last minute

lapinkon Cryptocurrency Exchange Locked Out of Funds After CEO's Death

I don’t get why would anyone carry that much crypto currencies on their laptop. Fake story or crooks ?