Test-Time Training (TTT) models context dependencies by adapting part of the model's weights (referred to as fast weights) during inference. This fast weight, akin to recurrent states in RNNs, stores temporary memories of past tokens in the current sequence. Existing TTT methods struggled to show effectiveness in handling long-context data, due to their inefficiency on modern GPUs. The TTT layers in many of these approaches operate with extremely low FLOPs utilization (often <5%) because they...| papers.cool
Self-attention performs well in long context but has quadratic complexity. Existing RNN layers have linear complexity, but their performance in long context is limited by the expressive power of their hidden state. We propose a new class of sequence modeling layers with linear complexity and an expressive hidden state. The key idea is to make the hidden state a machine learning model itself, and the update rule a step of self-supervised learning. Since the hidden state is updated by training ...| papers.cool
Computing the matrix square root or its inverse in a differentiable manner is important in a variety of computer vision tasks. Previous methods either adopt the Singular Value Decomposition (SVD) to explicitly factorize the matrix or use the Newton-Schulz iteration (NS iteration) to derive the approximate solution. However, both methods are not computationally efficient enough in either the forward pass or in the backward pass. In this paper, we propose two more efficient variants to compute ...| papers.cool
这篇文章我们来推导$\newcommand{msign}{\mathop{\text{msign}}}\msign$算子的求导公式。如果读者想要像《Test-Time Training Done...| spaces.ac.cn
在上文《msign算子的Newton-Schulz迭代（上）》中，我们试图为$\mathop{\text{msign}}$算子寻找更好的Newton-Schulz迭代，以期在有限迭代步数内能达到...| spaces.ac.cn
在之前的《Muon优化器赏析：从向量到矩阵的本质跨越》、《Muon续集：为什么我们选择尝试Muon？》等文章中，我们介绍了一个极具潜力、有望替代Adam的新兴优化器——“Muon”。随着相关研究...| spaces.ac.cn
随着LLM时代的到来，学术界对于优化器的研究热情似乎有所减退。这主要是因为目前主流的AdamW已经能够满足大多数需求，而如果对优化器“大动干戈”，那么需要巨大的验证成本。因此，当前优化器的变化，...| spaces.ac.cn
SVD（Singular Value Decomposition，奇异值分解）是常见的矩阵分解算法，相信很多读者都已经对它有所了解，此前我们在《低秩近似之路（二）：SVD》也专门介绍过它。然而，...| spaces.ac.cn
可能很多读者跟笔者一样，对矩阵的低秩近似有种熟悉而又陌生的感觉。熟悉是因为，低秩近似的概念和意义都不难理解，加之目前诸如LoRA等基于低秩近似的微调技术遍地开花，让低秩近似的概念在耳濡目染间就已...| spaces.ac.cn
事实上，除了写博客内容，在这几年里，笔者是花了相当一部分时间来做科学空间的“表面功夫”，为此还专门学了一点php、css和js。虽然不敢说精益求精，但总体来说网站的浏览体验应该比前几年要好得多。...| spaces.ac.cn