Coding & Communications

Stochastic representations fit iterative soft inference: LDPC decoding, belief propagation, and MIMO detection where messages are probabilities.

---

Why SC appears in decoders

graph LR
    CH[Channel LLRs] --> MSG[Soft messages]
    MSG --> SC[Bitstream / hybrid SC nodes]
    SC --> IT[BP / turbo-like iterations]
    IT --> DEC[Hard decisions]
    DEC --> BER[BER / FER]

• Check / variable node updates become tiny logic on wires that already mean probabilities.
• Fully parallel maps trade wires and SNGs for high throughput at low energy per bit when BER targets allow.

---

Design tensions

Tension	Comment
BER vs energy	Aggressive approximation can lift error floors
Bit-width accumulation	Hybrid recoding schemes manage growing precision
Mixed-signal CIM	MS-SCIM blends analog entropy with digital control

Background: Imaging, DSP & Coding.

---

📚 Primary literature (DOI only)

• Fully parallel stochastic LDPC decoder: DOI 10.1109/TVLSI.2024.3488716
• Stochastic BP iterative detection & decoding: DOI 10.1109/TVLSI.2024.3477963
• Mixed-signal SC-in-memory paradigm (MS-SCIM): DOI 10.1109/TCSI.2024.3501383
• Recoding hybrid stochastic numbers: DOI 10.1109/TCSI.2024.3492054

Domain browse: Paper Map → Coding & Communications.