HNHacker News
TopNewBestAskShowJobs

northlondoner

114 karma · joined March 22, 2023

Origins traced to North London, tech enthusiast.
submissionscomments
northlondoner··on Podcast: Solving Distributed Message Passing: NATS.io composite learning [video]
Good catch :) There is a typo, should read 'Message Passing'.
northlondoner··on Terence Tao: Machine Assistance and the Future of Research Mathematics [video]
Just a difference perspective on this. "Mathematician Joel David says current AI models are basically zero help for mathematics" see: https://x.com/slow_developer/status/2006909560007827686
northlondoner··on Podcast: Solving Distributed Message Passing: NATS.io composite learning [video]
Great podcast from Kris regarding NATS.io. If you come from MPI or Spark background for computations for composite computation, NATS seems like solves the inter-system communication.
northlondoner··on Can Quantum-Mechanical Description of Physical Reality Be Considered Complete? [pdf]
This is the original EPR paper, namely, Einstein–Podolsky–Rosen paradox. CERN's pdf. 1935 paper titled "Can Quantum-Mechanical Description of Physical Reality be Considered Complete?", Origin paper of 'quantum entanglement'.
northlondoner··on The concept of a distance over Riemannian Manifolds
Basic understanding of a metric tensor. For a leisurely read.
northlondoner··on Nats.io lightweight heterogeneous distributed, compositional, federaded learning
Looking into NATS.io as an infrastructure of choice in heterogeneous distributed learning and computing. JetStram concept with deep learning frameworks has a great potential. Lightweight is the key for this superb infrastructure technology.
northlondoner··on Nats.io JetStream with PyTorch: lightweight compositional learning
Lightweight solution to PyTorch for distributed training, across different modules: NATS.io provides a fault-tolarent communication bit across networks with publsh and fetch on given ids.
northlondoner··on The Wave Function of the Universe and Inflation
Probably in review now. arXiv physics is moderated with endorsements. Usually authors will update the Journal link on the arXiv. In this case the work is from a distinguished institution.
northlondoner··on Wigner Cat Phases: Transition to Quantum Chaos
A new tunable random matrix ensemble with intriguing properties.
northlondoner··on Julia vs. NumPy performance: Strategy for For-loop?
Interesting. Maybe this was due to matrix-matrix, matrix-vector and vector-vector separation in BLAS/LAPACK routines.
northlondoner··on The Wave Function of the Universe and Inflation
Usually, wavefunction from quantum mechanical sense is not used in classical mechanics in large scale. But in this context, how inflation occurs is an open problem in cosmology. A solid research from the distinguished National Observatory of Athens.
northlondoner··on New Yorker Dr. Berkan's New Channel RogoTRON
Thought provoking topics, from the nature of time, and information in biological and societal systems.
northlondoner··on Princeton-Affiliated Research: Exploring Singularity-Free Black Holes
Singularity-free black holes offer a more physically plausible alternative in strong gravity. This could offer potential insights into real astronomical observations, such as patterns in orbital data. This is a solid theoretical contribution in an active field.
northlondoner··on Julia vs. NumPy performance: Strategy for For-loop?
Thanks. Yes, we have novel use-cases which are not available out of box in NumPy.
northlondoner··on Lecture on spectral unfolding: Closely related to local noise removal
From the python package leymosun
northlondoner··on Functional programming and reliability: ADTs, safety, critical infrastructure
Yes.
northlondoner··on Julia vs. NumPy performance: Strategy for For-loop?
Thanks. numba is a good trade-off before moving to Julia completely.
northlondoner··on Julia vs. NumPy performance: Strategy for For-loop?
Some chat: Pyro/tf.prob are all more of an operations/stat research oriented tools, I gather but if there are some MC abstractions it might be good choice.

Yes, tested and used Jax for some prototyping but it felt like still has "two language problem". But Taichi looks quite interesting, will check out.

Of course, forgot to mention, it has to be lightweight, torch and tf are now huge platforms. Not sure, probably Julia-lang is much small.

northlondoner··on Origins of design of data frames: Software for data analysis (John Chambers)
Software for data analysis (John Chambers) is an R programming book but its unique position and style provides outstanding background on data frames and how statistical software should be. designed. One can of course extend this to AI software or data science or scientific computing.
northlondoner··on Memory Safety in C# vs. Rust
No. It is quite a viable cross-platform language and there is a large community. 1. C# works on Linux almost seamlessly. See the documentation: https://learn.microsoft.com/en-us/dotnet/core/install/linux 2. Actually C# specification is open. Meaning C# is like Java, anyone can implement in any platform. There are even alternating compilers, open sourced, thanks to Mono's efforts: https://www.mono-project.com
northlondoner··on Distributed and Compositional Learning Data-structure: C#-NATS example
This is quite underrated technology and its production level with linux comparability. If you come from Spark world this is RDDs on C#/NATS.

Brillant infrastructure production ready for compositional learning.

LINQ addition makes things easy to build data pipelines.

northlondoner··on Fault tolerance is key in compositional learning infrastructure: NATS and C#
Compositional learning usually means domain shifts in distributed training and learning that can be conceptually learn new concepts.

NATS provides a great fault tolerant messaging infrastructure in building compositional learning applications. C#'s distinction of being a cross-road among academic research, production system and business oriented language for applications makes it ideal tool. C#/NATS/LINQ combination provides amazing language primitives in building distributed business production systems.

northlondoner··on AI Takes over Boring Code: Is Software Engineering Losing Its Soul?
It is possible that 'GenAI' can help in generating boilerplate codes and repetitive tasks. But that's ironically just advanced automation.
northlondoner··on 'The biggest decision yet': Jared Kaplan on allowing AI to train itself
I think Jared is referring to "continuous training mode" (autonomous training). But that's itself open to interpretation. It sounds like old-good online learning.
northlondoner··on Ray Keynote: de-facto for compositional learning and a distributed heterogeneous [video]
Ray over internet is also possible as well as K8 https://docs.ray.io/en/latest/cluster/kubernetes/index.html
northlondoner··on Ray Keynote: de-facto for compositional learning and a distributed heterogeneous [video]
One thing valuable is "Fault Tolerance", https://docs.ray.io/en/latest/ray-core/fault-tolerance.html
northlondoner··on Ray Keynote: de-facto for compositional learning and a distributed heterogeneous [video]
HPC is very difficult domain to address all settings. With the current diverse hardware and compositional learning needs. Ray is now becoming de-facto framework for compositional learning infrastructure.
northlondoner··on Python is not a great language for data science
blubber, I think there might be some misconceptions. Just for the record.

R is not actually competing with those languages. R's design purpose is different. it is a general purpose computational language for scientists. There are FFIs (Foreign Function Interfaces) for all those languages.

R-Kotlin-Jave: https://journal.r-project.org/articles/RJ-2018-066/ R-Rust: https://cran.r-project.org/web/packages/using_rust.html R-C# : https://github.com/Open-Systems-Pharmacology/rsharp/

R is supporting C integration natively anyhow (see Chambers's book.

Regarding VB reference. VB was used in finance a lot to do some advanced maths. just a side remark.

northlondoner··on Python is not a great language for data science
Of course R isn't a complied language and probably not the same category as C/Rust as systems language but is not in the same category as VB. R is a serious scientific programming language used in non-trivial programs for industrial applications. See Posit's customers. I suggest John Chambers ( https://en.wikipedia.org/wiki/John_Chambers_(statistician) ) book, he explain how he designed S language, R's grandfather so to speak, Software for Data Analysis ( https://link.springer.com/book/10.1007/978-0-387-75936-4 ).
northlondoner··on Python is not a great language for data science
Has anybody else noticed how much Python took from Scala for type hints? I was using Scala around 2015 and when I see type hints, immediately recognise its similarity to Scala's approach.
← PreviousPage 2 of 4Next →