Exponentially Weighted Averages (C2W2L03)

Показать описание

DeepLearningAI

Рекомендации по теме

Комментарии

In this video we say: todaysEMA = (Beta * previousEMA) + ((1 - Beta) * todaysTHETA)

Wikipedia says: todaysEMA = (Beta * todaysTHETA) + ((1 - Beta) * previousEMA)

If we hold Beta at say .9 then these definitions would produce different results. However, if Wikipedia set Beta to .1 but this video set Beta to .9, the result would be equivalent.

If we wanted a reactionary measure, since this is essentially a large weighted average, I'd imagine that most of the time we want more weight on the new data point (todaysTHETA) and less weight on the previousEMA.

danieldeychakiwsky

There is a heading 'How to kill your grandmother' in one of Nassim Nicholas Taleb's book (Skin in the Game, if I remember correctly)) and it talks about distortion in the real picture created by averages ... according to Taleb keeping an eye on extreme values is far far more important, particularly the outliers, then the values which come closer to a so called average ..

DrOmarJavaid

It makes sense to initialize v0 to the first data point rather than zero. From the shown graph, it is obvious that he is doing it this way but he does not mention it

a.j.noushin

02:30 you are saying that if you compute this and plot it in red and then show the graph. I computed it and it doesn’t match up.
V1 = 4
V2 = 3.6 + 4.9 = 8.4 and so on.
When the temperature values are around 40 deg F.
Then how is the red graph alights with the temperatures ?

antonkorolev

I think, the period is aproximately calculated as 2/(1-beta).

vanheckcz

V(t) = (beta)V(t-1) + (1-beta) * @(t) or the other way. Python Panda dataframe implements the formula mentioned by me . Which is correct? weighted_average[0] = arg[0]; weighted_average[i] = + beta*arg[i].

saimadhavp

soo is this the thing people call "moving average" in context of neural networks to average weights over traininer iterations? If so that would mean you always average over 2 iterations? the current and the previous

RandomLogic

What is V? Is it (averaged) temperature with the same units as theta?

mailoisback

how is taking beta=0.9 equivalent to averaging temperature over 10 days?

sambitkumarbehura

If emwa = n-days MA then, why do we need it??

piyushsoni

says exponential.. but no exponents 🤔🤔🤔

KASANITEJ

“Exponentially weighted averages” is a quite pretentious name for the simplest _low pass filter_ there is.

leftaroundabout

Exponentially Weighted Averages (C2W2L03)

Exponentially Weighted Averages (C2W2L03)

Exponentially Weighted Moving Average or Exponential Weighted Average | Deep Learning

Understanding Exponentially Weighted Averages (C2W2L04)

Exponentially Weighted Moving Average or Exponential Weighted Average | Deep Learning

Bias Correction of Exponentially Weighted Averages (C2W2L05)

Exponential Weighted Average || Lesson 12 || Deep Learning || Learning Monkey ||

What are Exponentially Weighted Averages? | Moving Averages in Neural Networks

Exponentially Weighted Moving Average | How to calculate EWMA in Excel for A:C Ratio

Exponentially weighted averages

Exponentially Weighted Moving Average (EWMA) using Microsoft Excel

24a EWMA

76 Artificial Intelligence - Stochastic Gradient Descent , Exponentially Weighted Moving Average

Inside TimescaleDB | Time-weighted averages demo

When using the exponentially weighted moving average for purposes of forecasting...

EWMA

Deep Learning Module 2 Part 5/1: Exponentially Weight Average

#62 Pandas (Part 39): Calculate Exponential Weighted Mean, Variance and Std. in Python | Tutorial

#61 Pandas (Part 38): Intuition of Exponential Weighted Mean, Variance & Standard Dev. | Tutoria...

Term Sheet Economics: Weighted Average Anti-Dilution Provision

EWMA

065 EWMA Theory

Simple, Exponential and Weighted Moving Averages forex trading strategy

Exponentially Weighted RLS - Adaptive Filters - Advanced Digital Signal Processing

Averaging Weights Leads to Wider Optima and Better Generalization @ TWiML Online Meetup Americas