layer
AI
neural network inference tool
SHOW HN2019
251points
64comments
#136of 2,195 in 2019
Top 7%of 2019 by points
How HN took it
DebatedReal support and real criticism
11 positive, 12 neutral or asking, 1 mixed, 9 critical, out of 33 comments about the project. Thread score +0.09.
Idea: 5 for, 4 againstExecution: 1 for, 3 againstPerformance: 1 for, 2 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Unix pipeline idea is elegant and fun to explore
From 4 comments
Excerpt: "layer is a program for doing neural network inference the Unix way. Many modern neural network operations can be represented as sequential, unidirectional streams of…
peter_d_sherman on HNWhat's wonderful about this concept (and unix concept in general) is that the flexibility it gives you is amazing. You can for example pipe it over the network and distribute the…
mempko on HNI've been thinking about something like this for a long time, but could never quite wrap my head around a good way to do it (especially since I kept getting stuck on making it…
craftinator on HNGreat concept. Would like to see more of this idea applied to neural network processing and configuration in general (which in my experience can sometimes be a tedious,…
skvj on HN
Good for learning how neural networks work
From 1 comment
This might not be a great way to build neural networks (as other commenters have said regarding performance). But, it could be a great way to learn about neural networks. I…
xrd on HN
Scheme implementation is interesting
From 3 comments
Wonderful idea and the Chicken Scheme implementation looks nice also. I wrote some Racket Scheme code that reads Keras trained models and does inferencing but this is much…
mark_l_watson on HNIronically, I was only mildly interested in this before I read your comment and learned that it's in Scheme. Now I'm eager to check it out, haha.
seisvelas on HNThis, I'm primarily into Python, but the fact this is written in LISP makes it interesting alone.
rectangletangle on HN
Objections
Only inference, no training or backpropagation
From 3 comments
Great idea but however equally great caveat - it's just for (forward) inference. Unix pipelines are fundamentally one way and this approach won't work for back propagation.
luminati on HN> Now, that being said, there is one minor question I have, and that is, how would backpropagation apply to this apparently one-way model? The author mentioned that this is only…
kwaugh on HNThis kind of misses the point of the Unix philosophy of being able to dynamically reconfigure things - realistically, to get decent results, you'll need to do inference with the…
PeterisP on HN
Likely poor performance due to serializing data as text
From 2 comments
I don't see any claims about performance, but I would be very surprised if it was anything better than abysmal. In a modern neural network pipeline, just sending data to the CPU…
HALtheWise on HNBecause all the time spent marshalling and unmarshalling data structures into bags of textual bytes takes up CPU and does nothing but hasten the heat death of the universe. We…
bitwize on HN
Layers are tightly coupled and cannot be swapped freely
From 1 comment
My point was that this is not possible as the trained layers are intrinsically tightly coupled. You can't combine pre-trained sub networks in arbitrary manner without retraining.…
PeterisP on HN
Requests
Support for backpropagation
From 1 comment
I don’t see any reason you couldn’t just spit out the output and the derivative of the layer output with respect to the weights, then multiply and carry these all the way down.…
bigred100 on HN
Higher-level composition of pre-trained sub-networks
From 1 comment
I think that there is room for this idea at a higher level of abstraction where pre-trained sub networks are exposed as command line utilities. cat tweets.txt | layer…
iandanforth on HN
Questions
How would backpropagation work in a one-way pipeline
From 1 comment
Excerpt: "layer is a program for doing neural network inference the Unix way. Many modern neural network operations can be represented as sequential, unidirectional streams of…
peter_d_sherman on HN
Alternatives named
Clojure-based pipeline tools
From 1 comment
Clojure is ideal for Unix-style pipeline programming. Few people will use the command line for neural network inference, I think it would be better to use clojure to implement.…
lincpa on HN
Neural module networks from Berkeley
From 1 comment
See also Trevor Darrell's group's work on neural module networks: https://bair.berkeley.edu/blog/2017/06/20/learning-to-reason-with-neural-module-networks/
Donald on HN