tensorflow2.0 basic operations

Broadcasting (Radio)

• why？
  • concise
  • Provincial memory
• Example
•  

computation

outline

• +-*/
• **,pow,square
• sqrt
• //,%
• exp,log
• @ MATMUL
• linear layer

Merger and division

• tf.concat([],axis=)
  • merge
  • It does not create a new dimension
• tf.stack([],axis=)
  • merge
  • It creates a new dimension
• tf.unstack([],axis=)
  • And tf.stack () is reversible
• tf.split([],axis=,num_or_size_splits=[])
  • num_or_size_splits: into what

Statistics

• tf.norm ([], ord =, axis =) # tensor norm
  • Two norm: square and square root
  • A norm: absolute and
  • word: 1.2
• tf.reduce_min/max/mean
• tf.argmax/argmin
  • Returns the value of the index most
• tf.equal(a,b)
  • compared to
• tf.unique
  • Return two tensor

Sorting data

• black / argsort
  • tf.sort([],direction='')
  • tf.argsort: Back to index
• top_k
  • tf.math.top_k()
  • Applications: top-k accuracy
• top-5 acc.

Copy filling and

• tf.pad
  • Filling of data
  • tf.pad(a,[1,1],[1,1])
• tf.tile
  • Copy data
  • tf.tile(a,[2,2])
• broadcast_to

Tensor limiter

• clip_by_value
  • tf.maximum(a,2)
  • tf.minimum(a,8)
  • tf.clip_by_value(a,2,8)
• Relu
• clip_by_norm
• gradient

Higher-order OP

• where