Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2412.02078

A Simple Channel Compression Method for Brain Signal Decoding on Classification Task

Changqing Ji·Keisuke Kawasaki·Isao Hasegawa·Takayuki Okatani
View original source ↗Read source document ↗Export metadata ↓

Abstract

In the application of brain-computer interface (BCI), while pursuing accurate decoding of brain signals, we also need consider the computational efficiency of BCI devices. ECoG signals are multi-channel temporal signals which is collected using a high-density electrode array at a high sampling frequency. The data between channels has a high similarity or redundancy in the temporal domain. The redundancy of data not only reduces the computational efficiency of the model, but also overwhelms the extraction of effective features, resulting in a decrease in performance. How to efficiently utilize ECoG multi-channel signals is one of the research topics. Effective channel screening or compression can greatly reduce the model size, thereby improving computational efficiency, this would be a good direction to solve the problem. Based on previous work [1], this paper proposes a very simple channel compression method, which uses a learnable matrix to perform matrix multiplication on the original channels, that is, assigning weights to the channels and then linearly add them up. This effectively reduces the number of final channels. In the experiment, we used the vision-based ECoG multi-classification dataset owned by our laboratory to test the proposed channel selection (compression) method. We found that the new method can compress the original 128-channel ECoG signal to 32 channels (of which subject MonJ is compressed to 8 channels), greatly reducing the size of the model. The demand for GPU memory resources during model training is reduced by about 68.57%, 84.33% for each subject respectively; the model training speed also increased up around 3.82, 4.65 times of the original speed for each subject respectively. More importantly, the performance of the model has improved by about 1.10% compared with our previous work, reached the SOTA level of our unique visual based ECoG dataset

Explore related subjects

math.NAcs.NAq-bio.NC

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Changqing Ji, Keisuke Kawasaki, Isao Hasegawa, Takayuki Okatani. 2024-12-23. A Simple Channel Compression Method for Brain Signal Decoding on Classification Task. https://arxiv.org/abs/2412.02078

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

How many continuous measurements are needed to learn a vector?

One can recover vectors from $\mathbb{R}^m$ with arbitrary precision, using only $\lceil \log_2(m)\rceil +1$ continuous measurements that are chosen adaptively. This surprising result is explained and discussed, and we present applications to infinite-dimensional approximation problems.

math.NA↗
arXiv2026

IterativeCUR: Large Rank-Adaptive Approximation From a Small Recycled Sketch

The computation of accurate low-rank matrix approximations is central to improving the scalability of various techniques in machine learning, uncertainty quantification, and control. Traditionally, low-rank approximations are constructed using SVD-based approaches such as truncated SVD or Randomized SVD. Although these SVD approaches---especially Randomized SVD---have proven to be very computationally efficient, other low-rank approximation methods can offer even greater performance. One such approach is the CUR decomposition, which forms a low-rank approximation using direct row and column subsets of a matrix. Because CUR uses direct matrix subsets, it is also often better able to preserve native matrix structures like sparsity or non-negativity than SVD-based approaches and can facilitate data interpretation in many contexts. This paper introduces IterativeCUR, which draws on previous work in randomized numerical linear algebra to build a new algorithm that is highly competitive compared to prior work. IterativeCUR is adaptive in the sense that it takes as an input parameter the desired tolerance $ε$ and outputs (with arbitrarily high probability) an approximation of error bounded by $ε$, rather than requiring an a priori guess of the numerical rank. IterativeCUR typically runs significantly faster than both existing CUR algorithms and techniques such as Randomized SVD. Its asymptotic complexity is $\mathcal{O}(mn + (m+n)r^2)$ for an $m\times n$ matrix of output rank $r$. IterativeCUR relies on a single small sketch from the matrix that is successively downdated as the algorithm proceeds. We demonstrate through extensive experiments that IterativeCUR achieves up to $4\times$ speed-up over state-of-the-art pivoting-on-sketch approaches with no loss of accuracy, and up to $40\times$ speed-up over rank-adaptive randomized SVD approaches.

math.NA↗
arXiv2026

Multigrid with Linear Storage Complexity

As the discretization error for the solution of a partial differential equation (PDE) decreases, the precision required to store the corresponding coefficients naturally increases. Storing the solution's finite element coefficients explicitly requires $\mathcal O(n \log n)$ bits of storage, where $n$ is the number of degrees of freedom (DoFs). This paper presents a full multigrid method to compute the solution in a compressed format that reduces the storage complexity of the solution and intermediate vectors to $\mathcal O(n)$ bits. This reduction allows a matrix-free implementation to solve elliptic PDEs with an overall linear space complexity. For problems limited by the memory capacity of current supercomputers, we expect a memory footprint reduction of about an order of magnitude compared to state-of-the-art mixed-precision methods. We demonstrate the applicability of our algorithm by solving two model problems. Depending on the PDE and polynomial degree, but irrespective of the problem size, the solution vector on the finest grid requires between 4 and 12 bits per DoF, and the residual and correction require 3 to 6 bits each. Additional data is stored on the coarse grids with modestly increasing bit widths toward coarser grids.

math.NA↗