RT @RMBattleday: 📢📢Announcing this year's conference on the Mathematics of Neuroscience & AI (Split, Croatia, 27-30th May):

@SarthakChandra @sugsharma @FieteGroup TEM doesn't have any env specific learning for immediate appearance of grid cells in a new env. It's only hippocampal episodic memory binding of the grid code (the scaffold) to the particular sensory content of each env. Turn that off and grids will still be grids

@ThomasMiconi @NatureNeuro Definitely other existing models that "can quickly learn, and reconfigure, global relations from partial info (learn A>B, B>C, C>D, immediately know A>D)"!

@sugsharma @FieteGroup That's exactly what TEM does ;) @FieteGroup am I wrong?

@mindymagi Many thanks!

@ajaydiv @SuryaGanguli @dorrell_will @behrenstimb @GabyMohamady @NeuroCellPress Many thanks!

@hadivafaii Yes, a few of these ideas are speculated on in the discussion :)

@hadivafaii Yes they are conceptually related in that they are neural populations that like empty vessels can be filled with an arbitrary representation. The extra component is that they are connected together in such a way that allows structured sequence understanding

@apdemetriou Many thanks!

@dorrell_will @behrenstimb @GabyMohamady @SuryaGanguli Could be of interest to @dileeplearning @MillerLabMIT @criticalneuro @FieteGroup @nathanieldaw @tyrell_turing @GuangyuRobert @mattbotvinick @NeilBurgess10 and more