Crypto & Blockchain · Thu, Oct 29, 2026

Bitcoin, Ethereum, stablecoins, DeFi, smart contracts, scaling and consensus.

RSS feed
arXiv cs.LG · q-fin.PM · q-fin.ST

Sparse Autoencoders as Implicit Diversification Regularisers for Cross-Sectional Cryptocurrency Returns

In briefSparse autoencoders as a diversification tool for crypto returns

SummaryThis paper uses sparse autoencoders (SAEs) to describe daily returns of many cryptocurrencies with only a few active features. In a time-ordered evaluation against common benchmarks, SAE portfolios show consistently lower volatility, linked to lower correlations among selected assets. However, the SAE does not outperform a simple contrarian strategy on risk-adjusted returns, and its factor identities are unstable across training runs.

Weijia Han, Lisha Qu

arXiv cs.CR · cs.DC · cs.LG

DICE: Decentralized Incentive-driven Federated Learning for Efficient & Scalable Customization of Pretrained Models in Web3

In briefDICE: A decentralized, incentive-driven federated learning method for efficiently customizing pretrained models in Web3.

SummaryThe paper presents DICE, a blockchain-based federated learning algorithm for customizing large pretrained models among Web3 participants. It uses dynamic progressive learning and model slice verification so that participants unable to deploy the full model can still take part. It also introduces a tree-structured committee consensus protocol and an efficiency-driven incentive scheme to support more data holders. The authors report that theoretical analysis and experiments on vision and language models demonstrate its efficiency, effectiveness, and security.

Hong Lin, Zhongle Xie, Lidan Shou et al.