blob: f98490d9085e6ee9f4ea97bf756cdfbe6d10b8b3 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
|
# -*- coding: utf-8 -*-
#
# DPD Calculation Engine, heuristics we use to tune the parameters.
#
# http://www.opendigitalradio.org
# Licence: The MIT License, see notice at the end of this file
import numpy as np
def get_learning_rate(idx_run):
"""Gradually reduce learning rate from lr_max to lr_min within
idx_max steps, then keep the learning rate at lr_min"""
idx_max = 10.0
lr_min = 0.05
lr_max = 0.4
lr_delta = lr_max - lr_min
idx_run = min(idx_run, idx_max)
learning_rate = lr_max - lr_delta * idx_run / idx_max
return learning_rate
def get_n_meas(idx_run):
"""Gradually increase number of measurements used to extract
a statistic from n_meas_min to n_meas_max within idx_max steps,
then keep number of measurements at n_meas_max"""
idx_max = 10.0
n_meas_min = 10
n_meas_max = 20
n_meas_delta = n_meas_max - n_meas_min
idx_run = min(idx_run, idx_max)
learning_rate = n_meas_delta * idx_run / idx_max + n_meas_min
return int(np.round(learning_rate))
|
