An Adaptive Neural Network-Based Method for Tile Replacement in a Web Map Cache
0 likes662 views
This document presents an adaptive neural network-based method for tile replacement in web map caching, addressing the challenges of storage limitations and cache management. The proposed algorithm uses a feed-forward multilayer perceptron to predict tile request probabilities based on historical data from national tiled web map services in Spain. The results suggest that the neural network can effectively optimize cache replacements, supporting the efficient distribution of pre-generated map tiles.
![NEURAL NETWORK TARGET
During the training process, a training record is associated
with a boolean target (0 or 1) which indicates whether the
same tile is requested again or not in window
Once trained, the neural network output will be a real value
in the range [0,1] that must be interpreted as the probability
of receiving a successive request of the same tile within the
time window.](https://image.slidesharecdn.com/uva-iccsa11-110628093517-phpapp02/85/An-Adaptive-Neural-Network-Based-Method-for-Tile-Replacement-in-a-Web-Map-Cache-13-320.jpg)
An Adaptive Neural Network-Based Method for Tile Replacement in a Web Map Cache
- 1. AN ADAPTIVE NEURAL NETWORK-BASED METHOD FOR TILE REPLACEMENT IN A WEB MAP CACHE Ricardo García, Juan Pablo de Castro, María Jesús Verdú, Elena Verdú, Luisa M. Regueras and Pablo López Higher Technical School of Telecommunications Engineering University of Valladolid Santander, Spain, June 20-23, 2011
- 2. CONTENTS Web Map Services Tiled Map Services Web Proxy Cache Neural Network Cache Replacement Policy
- 3. WEB MAP SERVICES Clients Server GetMap parameters: Layers Styles GetCapabilities Coordinate Reference System Bbox GetMap Width Height Format Transparent BgColor Map images are generated on the fly Exceptions Tile flexible, but not scalable Elevation
- 4. TILED MAP SERVICES Bounding box and scales are Coarse constrained to discrete tiles resolution Level 0 Community specifications: WMS Tile Caching (WMS-C) Web Map Tile Service (WMTS) Detailed Propietary specifications: resolution Level l Microsoft Bing Maps Nasa World Wind Google Maps
- 5. PROXY WEB CACHE Clients Proxy Cache WMS Server GetCapabilities DataStore GetMap
- 6. BRUTE-FORCE APPROACH Caching the whole map can include millions of tiles Huge storage requirements ↑↑ Start-time to generate all the content Many GIS providers lack storage resources There are map services which update the cartography very often
- 7. PARTIAL CACHE When the cache runs out of space it is necessary to determine which tiles should be replaced by the new ones The cache replacement algorithm proposed in this work uses a neural network to estimate the probability that a request of a tile occurs before a certain period of time.
- 8. TRAINING DATA Trace requests from three public nation-wide tiled web map services in Spain: Cartociudad, IDEE-Base and PNOA. 1th – 7th March 2010 Cartociudad:25.922 reqs IDEE-Base: 94.520 reqs PNOA: 186.672 reqs Only cacheable requests considered
- 9. HEATMAPS
- 11. NEURAL NETWORK PARAMETERS Parameter Value Architecture Feed-forward Multilayer Perceptron Inputs 3 (recency, frequency, size) Hidden layers 2 Neurons per hidden layer 3 Output 1 (probability of a future request) Activation functions Log-sigmoid in hidden layers, Hyperbolic tangent sigmoid in output layer Error function Minimum Square Error (mse) Training algorithm Backpropagation with momentum Learning method Supervised learning Weights update mode Batch mode Learning rate 0.05 Momentum constant 0.2
- 13. NEURAL NETWORK TARGET During the training process, a training record is associated with a boolean target (0 or 1) which indicates whether the same tile is requested again or not in window Once trained, the neural network output will be a real value in the range [0,1] that must be interpreted as the probability of receiving a successive request of the same tile within the time window.
- 14. SIMULATION RESULTS IDEE-Base Cartociudad PNOA
- 16. CONCLUSIONS Serving pre-generated map image tiles from a server-side cache has become a popular way of distributing map imagery on the Web Storage needs are often prohivitive which forces the use of partial caches A feed-forward multilayer perceptron can effectively take replacement decisions based on recency, frequency and size of map requests
- 17. THANK YOU! ricgar@tel.uva.es