Unbiased approximations of products of expectations
40 mins 28 secs,
123.32 MB,
WebM
640x360,
29.97 fps,
44100 Hz,
416.06 kbits/sec
Share this media item:
Embed this media item:
Embed this media item:
About this item
| Description: |
Lee, A
Tuesday 4th July 2017 - 09:45 to 10:30 |
|---|
| Created: | 2017-07-21 12:33 |
|---|---|
| Collection: | Scalable inference; statistical, algorithmic, computational aspects |
| Publisher: | Isaac Newton Institute |
| Copyright: | Lee, A |
| Language: | eng (English) |
| Distribution: |
World
(downloadable)
|
| Explicit content: | No |
| Aspect Ratio: | 16:9 |
| Screencast: | No |
| Bumper: | UCS Default |
| Trailer: | UCS Default |
| Abstract: | I will describe recent work with Simone Tiberi (Zurich) and Giacomo Zanella (Bocconi), on the unbiased approximation of a product of n expectations. Such products arise, e.g., as values of the likelihood function in latent variable models, and unbiased approximations can be used in a pseudo-marginal Markov chain to facilitate inference. A straightforward, standard approach consists of approximating each term using an independent average of M i.i.d. random variables and taking the product of these approximations. While simple, this typically requires M to be O(n) so that the total number of random variables required is N = Mn = O(n^2) in order to control the relative variance of the approximation. Using all N random variables to approximate each expectation is less wasteful when producing them is costly, but produces a biased approximation. We propose an alternative to these two approximations that uses most of the N samples to approximate each expectation in such a way that the estimate of the product of expectations is unbiased. We analyze the variance of this approximation and show that it can result in N = O(n) being sufficient for the relative variance to be controlled as n increases. In situations where the cost of simulations dominates overall computational time, and fixing the relative variance, the proposed approximation is almost n times faster than the standard approach to compute. |
|---|---|
Available Formats
| Format | Quality | Bitrate | Size | |||
|---|---|---|---|---|---|---|
| MPEG-4 Video | 640x360 | 1.94 Mbits/sec | 589.21 MB | View | Download | |
| WebM * | 640x360 | 416.06 kbits/sec | 123.32 MB | View | Download | |
| iPod Video | 480x270 | 522.15 kbits/sec | 154.70 MB | View | Download | |
| MP3 | 44100 Hz | 249.73 kbits/sec | 74.08 MB | Listen | Download | |
| Auto | (Allows browser to choose a format it supports) | |||||