Robert Lucky's 1965 adaptive equalizer broke the telephone-line 'distortion barrier' using the same steepest-descent adaptive-filter logic as Widrow's LMS

Widrow & Lehr's 1990 retrospective (the source behind claim-madaline-rule-i-first-layer-trainable-second-fixed) credits "outstanding work by Lucky and others at Bell Laboratories" for turning adaptive-filter theory into a commercial telecommunications breakthrough — a tangent that leads away from neural-net history into modem engineering.

Claim 1 (mechanism, Tier 1, primary): Voice telephone lines in the early 1960s were capped at roughly 2400 bits/second by intersymbol interference, not noise. Lucky's own 1965 paper states: "Present data rates on voice telephone channels are limited to about 2400 bits per second... the nonuniform transmission characteristics of the channel cause what might be termed a distortion barrier, prohibiting faster transmission." (Lucky 1965, Bell System Technical Journal 44(4), p.548)

Claim 2 (mechanism, Tier 1, primary): His fix — the automatic/adaptive equalizer — used a tapped delay line (transversal filter) whose tap gains were set by a steepest-descent minimization during a training period of known test pulses: "An easily implemented system for automatic equalization is described which makes use of a steepest-descent technique of minimization. The equalizer is automatically set prior to data transmission in a training period during which a series of test pulses is transmitted." (same source, p.547) This is structurally the same adaptive-filter logic as Widrow's LMS algorithm, applied to a different physical channel.

Claim 3 [unverified-quant — needs primary]: Secondary sources (Wikipedia, ETHW-derived bios) report the equalizer's commercial deployment enabled a jump to 9600 bps — 4x the prior ceiling — but this figure was not located in the primary 1965 paper itself and needs a primary confirmation before it's load-bearing.

Why this was hop-worthy

Same adaptive-filter mathematics (steepest descent on a tapped delay line) solved an entirely different 1960s bottleneck — telephone-line data rates — independently of and parallel to Widrow's neural-net line of work, making it a clean lateral hop out of ML history into telecom engineering.

Further leads

• Lucky's later 1966/1967 BSTJ follow-ups on general-purpose automatic equalizers.
• Whether Lucky and Widrow's groups were aware of each other's work in the mid-1960s, or converged independently.