liamdgray

7 ms·

1.
by liamdgray 5mo ago
Great idea. I would love to see your transpiler Mind sharing it?
2.
by liamdgray 6mo ago
Please do!
3.
by liamdgray 10mo ago
Some students even wish for a ban to reduce the pressure to keep up with social media. That reminded me of Warren Buffet asking for his kind and to be taxed more.
4.
by liamdgray 1y ago
Abstract: "We introduce an Invertible Symbolic Regression (ISR) method. It is a machine learning technique that generates analytical relationships between inputs and outputs of a given dataset via invertible maps (or architectures). Th
5.
ISR: Invertible Symbolic Regression (2024)(arxiv.org)
7 points by liamdgray 1y ago|1 comments
6.
by liamdgray 1y ago
Abstract: "Symbolic equations are at the core of scientific discovery. The task of discovering the underlying equation from a set of input-output pairs is called symbolic regression. Traditionally, symbolic regression methods use hand-
7.
Neural Symbolic Regression that scales (2021)(proceedings.mlr.press)
1 points by liamdgray 1y ago|1 comments
8.
by liamdgray 1y ago
Abstract: "In recent years, self-attention has become the dominant paradigm for sequence modeling in a variety of domains. However, in domains with very long sequence lengths the O(T^2) memory and O(T^2H) compute costs can make using t
9.
Recasting Self-Attention with Holographic Reduced Representations (2023)(proceedings.mlr.press)
2 points by liamdgray 1y ago|1 comments
10.
by liamdgray 1y ago
Abstract: "Contemporary large models often exhibit behaviors suggesting the presence of low-level primitives that compose into modules with richer functionality, but these fundamental building blocks remain poorly understood. We invest
11.
Composing Linear Layers from Irreducibles(arxiv.org)
2 points by liamdgray 1y ago|1 comments
12.
by liamdgray 1y ago
Abstract: "Current automated systems have crucial limitations that need to be addressed before artificial intelligence can reach human-like levels and bring new technological revolutions. Among others, our societies still lack level-5
13.
Introduction to latent variable energy-based models (2024)(iopscience.iop.org)
2 points by liamdgray 1y ago|1 comments
14.
by liamdgray 1y ago
Abstract: "Associative Memories like the famous Hopfield Networks are elegant models for describing fully recurrent neural networks whose fundamental job is to store and retrieve information. In the past few years they experienced a su
15.
Modern Methods in Associative Memory(arxiv.org)
5 points by liamdgray 1y ago|1 comments
16.
by liamdgray 1y ago
"Abstract A model of associative memory is studied, which stores and reliably retrieves many more patterns than the number of neurons in the network. We propose a simple duality between this dense associative memory and neural networks
17.
Dense Associative Memory for Pattern Recognition (2016)(proceedings.neurips.cc)
2 points by liamdgray 1y ago|1 comments
18.
by liamdgray 1y ago
Is this intended to run on a quantum computer? You mention Grover's search algorithm, for example.
19.
by liamdgray 1y ago
Recorded at the ML in PL 2019 Conference, the University of Warsaw, 22-24 November 2019. Anton Osokin (Higher School of Economics, Moscow), https://aosokin.github.io/ Slides available at https://docs.mlinpl.org&#
20.
How to put algorithms into neural networks? (2019) [video](youtube.com)
3 points by liamdgray 1y ago|1 comments
21.
by liamdgray 1y ago
Abstract Tomographic Synthetic Aperture Radar (TomoSAR) building object height inversion is a sparse reconstruction problem that utilizes the data obtained from several spacecraft passes to invert the scatterer position in the height direct
22.
Improved Analytic Learned Iterative Shrinkage Thresholding Algorithm(mdpi.com)
2 points by liamdgray 1y ago|1 comments
23.
by liamdgray 1y ago
Keywords: associative memory, Hopfield networks, transformers, attention, in-context learning, denoising Abstract: We introduce in-context denoising, a task that refines the connection between attention-based architectures and dense associa
24.
In-context denoising with one-layer transformers(openreview.net)
3 points by liamdgray 1y ago|1 comments
25.
by liamdgray 1y ago
Abstract: "Deep neural networks provide unprecedented performance gains in many real world problems in signal and image processing. Despite these gains, future development and practical deployment of deep networks is hindered by their
26.
Algorithm Unrolling: Interpretable, Efficient Deep Learning for Sig&Img (2019)(arxiv.org)
1 points by liamdgray 1y ago|1 comments
27.
Hopfield Networks Is All You Need (2020)(arxiv.org)
41 points by liamdgray 1y ago|6 comments
28.
by liamdgray 1y ago
In Section 2.8, you write "full implementation details and extended results are provided in the appendix." Which appendix? I imagine you may be withholding some of the details until after the conference at which, it seems, you wil
29.
by liamdgray 1y ago
I ran across this paper because the recent "subliminal learning" results reminded me of holography. So I asked o4-mini-high to explore potential relationships. It lead me to this. https://chatgpt.com/share/68
30.
by liamdgray 1y ago
Abstract: "Large Language Models (LLMs) exhibit remarkable capabilities but suffer from apparent precision loss, reframed here as information spreading. This reframing shifts the problem from computational precision to an information-t

More ›