loss less compression algorithms take an input stream of raw data and are able to represent that stream in a reduced form. The key point, is the compress data can be exploded back to it's original form exactly. ZIP, PNG are examples of every day routines, but there's lots of lots of them there !
lossy compression algo's are conceptually the same idea, they take some input data, and work out a way to represent that data in less bytes. The key difference in lossy decompression is it's not imperative that the output data be an perfect match of the original input. Making them not suitable for packing EXE's files or an text file, but can be very effective with Images / Video / Sound & data sets. (JPG,Mp3 etc etc)
