Implement Tensor

[CameronMatthew]

What should a tensor allow:

• creation from a flat array with a shape
• creation of identity
• creation of zeros
• creation of ones
• scalar operations
  • multiply
  • add
  • divide
  • subtract
  • power (including fractional power)
  • ~logical operations (>, <, == etc)~
• element-wise tensor operations (between two tensors of the same shape)
  • multiply
  • add
  • divide
  • subtract
  • power (including fractional)
  • ~logical operations~
• creation from tile of a smaller tensor
• slice a sub-tensor from an existing tensor
• iterate over one axis
• reduction operations (one axis)
  • sum
  • mean
  • median
  • mode
  • max
  • min
  • stddev
• collapse operations (e.g. multiply every element in tensor together)
  • sum
  • multiply
• tensor product

[siliconlad]

Would the logical operators return a Boolean tensor?

[CameronMatthew]

Would the logical operators return a Boolean tensor?

Didn't think about that lol. I guess so. But that seems like an extra type that we just don't need right now. Let's leave those out until we absolutely need them.

[siliconlad]

I suppose you could also have a tensor which has 1's and 0's for true and false respectively. So might not be too difficult to implement, but still. Probably best to wait until we need it.