Hi,
I think that one possible alternative to (securely) accelerate blocks creation is to combine the following proposals:
http://www.tik.ee.ethz.ch/file/49318d3f56c1d525aabf7fda78b23fc0/P2P2013_041.pdf and
http://www.mail-archive.com/bitcoin-development@lists.sourceforge.net/msg03307.html. Basically, the information-propagation paper propose to introduce special kind of messages, that deliver only the block header (including the nonce) so that it will still be possible to validate the PoW (and therefore DoS attacking the system by spamming fake headers is not possible), in order to make blocks propagate faster. Peter Todd suggests allowing blocks to contain invalid transactions, e.g. double spending of money that was already spent in some earlier block. Adopting the first spending transaction and ignoring all letter attempts is enough to maintain security.
So if a miner founds a new block and separately propagates the block header and the block itself (or even only a list of the transactions' hashes), propagation of the block header will quickly inform all the network of the new founded block, and miners can start mining on top of that block even if they have not received yet the block itself. In the worst case, the second miner might include transactions that were already included in the previous block or other invalid transactions (of whom he/she we get no fees).
Yet a possible problem with that solution is regarding the Bloom filter technique, which doesnt work well if invalid transactions are included in the blocks. So I would like to propose here a possible solution for that, by punishing double spenders: each transaction must include a minimal deposit that is unlocked only after a few blocks (for example, 10) and only in case no earlier double spending transaction have been found. However if there has been such an earlier transaction, the deposit money will be destroyed (may be part of it can be used to compensate the miner who mined the block that contains this invalid transaction). Obviously, if the same exact transaction has been included twice within different blocks in the chain, it will not be regarded as a double-spending transaction.
I personally much prefer the GHOST proposal, and think it will be more efficient.
Lear.