diff options
author | Soumith Chintala <soumith@gmail.com> | 2016-09-02 21:43:12 +0300 |
---|---|---|
committer | GitHub <noreply@github.com> | 2016-09-02 21:43:12 +0300 |
commit | a81143f095c2aeb897f3f043699d10d0c4a96375 (patch) | |
tree | 9af3e0463f4d71104d3620abc12a2bf67fe2cff5 | |
parent | 40d3a23d006e89c97a29a6714efbde9e38c89ad4 (diff) |
fix bilinear doc
-rw-r--r-- | doc/simple.md | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/doc/simple.md b/doc/simple.md index 3058c5e..302e4d8 100644 --- a/doc/simple.md +++ b/doc/simple.md @@ -139,7 +139,7 @@ module = nn.Bilinear(inputDimension1, inputDimension2, outputDimension, [bias = ``` Applies a bilinear transformation to the incoming data, i.e. `\forall k: y_k = x_1 A_k x_2 + b`. The `input` tensor given in `forward(input)` is a table containing both inputs `x_1` and `x_2`, which are tensors of size `N x inputDimension1` -and `N x inputDimension1`, respectively. The layer can be trained without biases by setting `bias = false`. +and `N x inputDimension2`, respectively. The layer can be trained without biases by setting `bias = false`. You can create a layer in the following way: @@ -1479,4 +1479,4 @@ Selects the highest `k` values for each feature in the feature map sequence prov If `factor` is not provided, `k = minK`, else the value of k is calculated with: ```lua k = math.max(minK, math.ceil(factor*nInputFrame))) -```
\ No newline at end of file +``` |