class: monash-bg-blue center middle hide-slide-number <div class="bg-black white" style="width:45%;right:0;bottom:0;padding-left:5px;border: solid 4px white;margin: auto;"> <i class="fas fa-exclamation-circle"></i> These slides are viewed best by Chrome and occasionally need to be refreshed when not loading properly. <!-- See here for <a href=index.pdf>PDF <i class="fas fa-file-pdf"></i></a>. --> </div> <br> .white[Press the **right arrow/space** to progress to the next slide!] --- count: false background-image: url(images/anzsc2021logo.png) background-size: cover class: hide-slide-number title-slide .item.center[ # <span style="text-shadow: 2px 2px 50px white;">Manifold Learning with<br/>Approximate Nearest Neighbors</span> ] .center.shade_monash_blue.animated.bounceInUp.slower[ <!-- <br><br> --> <!-- ## <span style="color: #ccf2ff; text-shadow: 10px 10px 100px white;"></span> --> <br> Presented by Fan Cheng (Monash University) Rob J Hyndman (Monash University)<br/>Anastasios Panagiotelis (University of Sydney) Department of Econometrics and Business Statistics <img src="images/monash-one-line-reversed.png" style="width:500px">
<i class="fas fa-envelope faa-float animated "></i>
Fan.Cheng@monash.edu
<i class="fab fa-twitter faa-float animated faa-fast "></i>
@fanchengfc <br><br> .bottom_abs.width100.bg-black[ Australian and New Zealand Statistical Conference (ANZSC) July 6, 2021 ] ] ??? <div class="grid-row" style="grid: 1fr / 2fr;"> .item.center[ # <span style="text-shadow: 2px 2px 30px white;">Manifold Learning with<br/>Approximate Nearest Neighbors</span> <!-- # .monash-blue.outline-text[Manifold Learning with<br/>Approximate Nearest Neighbors] --> ## <span style="color:;text-shadow: 2px 2px 30px black;"></span> ] </div> <style> p.caption { font-size: 0.8em; } </style> --- # Outline - Motivation - Manifold learning - Approximate nearest neighbors - Embedding quality measures - MNIST data results - Application to smart meter data - Conclusions --- class: split-two # Irish Smart Meter Data .row[ .split-five[ .row[ ] .row[.content[ - Problem of interest: electricity usage patterns of households ]] .row[.content[ - Typical/anomalous households in distributions ]] .row[.content[ - Half-hourly data for 535 days in 3639 households ]] .row[.content[ - Empirical discrete distributions: `\(48 \times 7=336\)` distributions per household ]] ] ] .row[ .split-two[ .column[