How do we process audio data with fully connected network?

[delta2323]

@soumith wrote we will use MSR's 5-layered fully connected network for evaluating frameworks for audio use case. But how do we use this network for processing time-series data?

One obvious way is not to make lateral connections (i.e. connection along time dimension) and to process data point of each time independently with the network. But I do not know this is the standard way.

Let me know if I miss some information. Thanks in advance.

[soumith]

cc: @Alexey-Kamenev