Datasets:

zai-org
/

RPC-Bench

@@ -217,7 +217,7 @@ Input: training set $X$ ; basis $V = [ \pmb { v } _ { 1 } , \dots , \pmb { v } _
 Initialize: noise variances $\Sigma _ { 0 } = \mathrm { { D i a g } } [ \sigma _ { 1 , 0 } , . . . , \sigma _ { D , 0 } ]$ .
 Output: robust network $f _ { \boldsymbol { \theta } , \Sigma } ^ { \mathrm { S N } }$ , noise variances $\Sigma _ { T } = \operatorname { \bar { D i a g } } [ \sigma _ { 1 , T } ^ { 2 } , \dots , \sigma _ { D , T } ^ { 2 } ]$ .
 1: for epoch $t = 1 \dots T$ do
-2: for mini-batch $B = \{ { \pmb x } _ { 1 } , \ldots , { \pmb x } _ { r } \}$ do θ ← BASE\`∞ f SNθ,Σt {xi}ri=1, θ . BASE() Training
 3: end for
 4: if $t$ mod $U _ { f } = 0$ then . SNAP Distribution Update once every $U _ { f }$ epochs
 5: for mini-batch $B = \{ { \pmb x } _ { 1 } , \ldots , { \pmb x } _ { r } \}$ do

 Initialize: noise variances $\Sigma _ { 0 } = \mathrm { { D i a g } } [ \sigma _ { 1 , 0 } , . . . , \sigma _ { D , 0 } ]$ .
 Output: robust network $f _ { \boldsymbol { \theta } , \Sigma } ^ { \mathrm { S N } }$ , noise variances $\Sigma _ { T } = \operatorname { \bar { D i a g } } [ \sigma _ { 1 , T } ^ { 2 } , \dots , \sigma _ { D , T } ^ { 2 } ]$ .
 1: for epoch $t = 1 \dots T$ do
+2: for mini-batch $B = \{ { \pmb x } _ { 1 } , \ldots , { \pmb x } _ { r } \}$ do θ ← BASE\`∞ f SNθ,Σt {xi}ri=1, θ . BASE() Training
 3: end for
 4: if $t$ mod $U _ { f } = 0$ then . SNAP Distribution Update once every $U _ { f }$ epochs
 5: for mini-batch $B = \{ { \pmb x } _ { 1 } , \ldots , { \pmb x } _ { r } \}$ do