On Tue, Sep 4, 2012 at 7:50 PM, Ben Goertzel <[email protected]> wrote: > Well it's a subtle issue... > > For prediction of real-valued time series, one uses a variety of different > accuracy metrics, often custom ones based on what one is trying to use the > predictions for... > > Presumably predictions according to custom accuracy metrics, could be > mapped into lossy compressors with different metrics for lossiness.... But > the formulation of an approximate predictor in terms of lossy compression > may sometimes be a lot uglier...
For lossless compression, you run the prediction algorithm in both the compressor and decompresser and you encode the prediction error. Generally the smaller the error, the better the compression. -- Matt Mahoney, [email protected] ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
