Learning-to-Transition for Large-scale and High-Order MIMO Detection
A paper proposes a learning-to-transition (L2T) framework for high-order MIMO detection, using a channel-coupled Transformer and blockwise autoregressive factorization. Hard-output detection uses a transition network trained with a residual-to-BER curriculum. Soft-output detection clones the hard policy into an untied soft-input soft-output iterative detection and decoding (IDD) receiver.
The paper develops a learning-to-transition (L2T) framework for high-order multiple-input multiple-output (MIMO) detection. It formulates detection as a stochastic sequence of complete-vector transitions, where a channel-coupled Transformer updates instance embeddings and sampling policy, and blockwise autoregressive factorization captures inter-stream dependence. Hard-output detection applies a transition network recursively, trained via a residual-to-BER curriculum that first learns search geometry from the exact residual metric and then aligns with transmitted-bit accuracy. For soft-output reception, the trained hard policy is cloned at the parameter level into every layer of an untied soft-input soft-output iterative detection and decoding (IDD) receiver, enabling layer- and round-specific specialization under decoder feedback.
The L2T framework combines a channel-coupled Transformer with blockwise autoregressive factorization to manage high-order MIMO search complexity. The residual-to-BER curriculum training first optimizes for residual metric geometry, then fine-tunes for bit error rate. Parameter-level cloning from hard to soft detection preserves zero-prior search dynamics while allowing untied layers to specialize per IDD round.
This approach could improve MIMO detection performance in 5G/6G base stations and high-throughput wireless systems, potentially reducing computational load while maintaining soft information quality for channel decoding. Adoption would require validation on real channel models and integration with existing receiver architectures.
Improved MIMO detection can enhance spectral efficiency and reliability in wireless communications, benefiting telecom equipment vendors and network operators. The method may reduce processing latency or power consumption in baseband units, offering cost savings in dense deployments.
Next signals include publication of peer-reviewed results, benchmark comparisons against state-of-the-art MIMO detectors, and potential follow-up work on hardware implementation or extension to other wireless standards.