next up previous
Next: Choosing an Action Up: Memory Model Previous: Storing to Memory

Retrieving from Memory


Since individual training examples affect multiple memory locations, we use a simple technique for retrieving tex2html_wrap_inline456 from memory when deciding whether to shoot or to pass. We round tex2html_wrap_inline252 to the nearest tex2html_wrap_inline662 for which Mem[ tex2html_wrap_inline664 ] is defined, and then take tex2html_wrap_inline666 as the value of tex2html_wrap_inline456 . Thus, each Mem[ tex2html_wrap_inline670 ] represents tex2html_wrap_inline456 for tex2html_wrap_inline674 . Notice that retrieval is much simpler when using this technique than when using kNN or kernel regression: we look directly to the closest fixed memory position, thus eliminating the indexing and weighting problems involved in finding the k closest training examples and (possibly) scaling their results. We used this retrieval technique throughout our experiments, concentrating the trickiness of our function learning at storage time.

Peter Stone
Mon Dec 11 15:42:40 EST 1995