When designing a deep learning model, whenever a decision must be made, the answer is simple: multiply your data by a weight matrix. For example, in an LSTM, need to forget some data? Just multiply by a weight matrix to get the forget gate. The system will learn what you meant.
5
31
3
265
It's remarkable that so much power comes from literally just multiplying by a matrix.

Dec 5, 2018 · 9:56 PM UTC

6
6
1
92
Replying to @gdb
Gives us insight into the multidimensional nature of human cognition; We suffered dearly under the unidimensionalism of behaviorism. Symbol processing mindset was an improvement but still prayed at the church of mechanistic thinking. Now we can move into #holisticai
Replying to @gdb
Nature does so much with so few operators - why shouldn't we!
Replying to @gdb
And, at the core of most every digital computer are just billions upon billions of NAND gates.
3
Replying to @gdb
A weight matrix in a human readable form is just a way to express which parameters are important and by how much.
Replying to @gdb
it's remarkable that many problems are amenable - historically it is probably not uncommon that the complexity of a problem is overestimated until a solution is found