Skip to content

Instantly share code, notes, and snippets.

@holianh
holianh / ml_dl_scenarios.md
Created August 1, 2018 02:18 — forked from dusenberrymw/ml_dl_scenarios.md
Interesting Machine Learning / Deep Learning Scenarios

Interesting Machine Learning / Deep Learning Scenarios

This gist aims to explore interesting scenarios that may be encountered while training machine learning models.

Increasing validation accuracy and loss

Let's imagine a scenario where the validation accuracy and loss both begin to increase. Intuitively, it seems like this scenario should not happen, since loss and accuracy seem like they would have an inverse relationship. Let's explore this a bit in the context of a binary classification problem in which a model parameterizes a Bernoulli distribution (i.e., it outputs the "probability" of the true class) and is trained with the associated negative log likelihood as the loss function (i.e., the "logistic loss" == "log loss" == "binary cross entropy").

Imagine that when the model is predicting a probability of 0.99 for a "true" class, the model is both correct (assuming a decision threshold of 0.5) and has a low loss since it can't do much better for that example. Now, imagine that the model

@holianh
holianh / clean_audio.sh
Last active July 5, 2020 22:25 — forked from devoncrouse/clean_audio.sh
Using Sox (http://sox.sourceforge.net) to remove background noise and/or silence from audio files (individually, or in batch).
# Create background noise profile from mp3
/usr/bin/sox noise.mp3 -n noiseprof noise.prof
# Remove noise from mp3 using profile
/usr/bin/sox input.mp3 output.mp3 noisered noise.prof 0.21
# Remove silence from mp3
/usr/bin/sox input.mp3 output.mp3 silence -l 1 0.3 5% -1 2.0 5%
# Remove noise and silence in a single command
@holianh
holianh / tensorflow_log_loader.py
Created January 10, 2019 01:49 — forked from tomrunia/tensorflow_log_loader.py
Reading out binary TensorFlow log file and plotting process using MatplotLib
import numpy as np
from tensorflow.python.summary.event_accumulator import EventAccumulator
import matplotlib as mpl
import matplotlib.pyplot as plt
def plot_tensorflow_log(path):
# Loading too much data is slow...
tf_size_guidance = {