学习视频
1.tf.norm
Here talks about Vector Norm
a=tf.ones([2,2])
tf.norm(a) #numpy=2.0 (1^2+1^2+1^2+1^2)^0.5=2.0
tf.sqrt(tf.reduce_sum(tf.square(a))) #numpy=2.0
a=tf.ones([4,28,28,3])
tf.norm(a) #numpy=96.99484
tf.sqrt(tf.reduce_sum(tf.square(a))) #numpy=96.99484
2.L1 Norm
b=tf.ones([2,2])
tf.norm(b,ord=2,axis=1) #numpy=array([1.4142135,1.4142135])
tf.norm(b,ord=1) #numpy=4.0 1+1+1+1=4
tf.norm(b,ord=1,axis=0) #numpy=array([2.,2.])
tf.norm(b,ord=1,axis=1) #numpy=array([2.,2.])
3.reduce_min/max/mean
a=tf.random.normal([4,10])
tf.reduce_min(a),tf.reduce_max(a),tf.reduce_mean(a) #numpy=-1.1872,numpy=2.135,numpy=0.352
tf.reduce_min(a,axis=1) #numpy=array[-0.393,-1.423.-1.093,-1.136]
4.argmax/argmin
a=tf.random.normal([4,10])
tf.argmax(a).shape #TensorShape([10])
tf.argmax(a) #numpy=array([0,0,2,3,1,5,7,0,3,4])
tf.argmin(a).shape #TensorShape([10])
5.tf.equal
a=tf.constant([1,2,2,5,4])
b=range(5)
b #[0,1,2,3,4]
tf.equal(a,b) #numpy = array([False,False,True,False,True])
res=tf.equal(a,b)
tf.reduce_sum(tf.cast(res,dtype=tf.int32)) #numpy=2
6.Accuracy
a=np.array([[0.1,0.2,0.7],[0.9,0.02,0.08]])
y=np.array([1,0])
pred = tf.cast(tf.argmax(a,axis=1),dtype=int32) #numpy=array([2,0])
tf.equal(y,pred) #numpy=array([False,True])
correct=tf.reduce_sum(tf.cast(tf.equal(y,pred),dtype=int32)) #numpy=1
accuracy=correct/len(y) #numpy=0.5
7.tf.unique
a=tf.constant([4,2,2,4,3])
tf.unique(a) #numpy=array([0,1,1,0,2])