* Add the difference between training and predicting under Autograd

* fix training and predicting

* fix how-to-use.md

* fix how-to-use.md

* fix install.md

* Update install.md

1.有不少语病的地方作了修正：
如：“所以在SoftmaxCrossEntropyLoss里加入了额外**了**axis=1选项。”中“了”应为“的”。
2.表述模糊的地方增加了描述：
如：“我们可以得到和x一样的输出。”转置卷积只是得到了形状一致的输出，并不能说与x一样，故增加了“形状”在x之后。
3.术语翻译改为了更常见的：
“双线性差值”一般译为“双线性插值”

* fix how-to-use.md

* fix how-to-use.md

* fix gur.md

* fix lstm.md

* '修改笔误，应该是根据输入计算出输出'

* '修改笔误'

* '修改笔误'

* 去掉了重复语病并对句子进行了润色

* 润色了一丢丢

“第一只狗”被写成“地一只狗”了

* fix how-to-use.md

* fix how-to-use.md

* fix typo in transpose-conv intro in section9.9.1

* Update fcn.md

* fix typo in conv-layer.md

* fix typo in padding-and-strides.md

* fix typo in channels.md

* fix typo in deferred-init.md

* fix typo in lenet.md

* fix typo in alexnet.md

* fix typo in vgg.md

* fix typo in nin.md

* fix typo in batch-norm.md

* fix typo in resnet.md

* '修改笔误，应该是根据输入计算出输出'

* '修改笔误'

* '修改笔误'

* fix spelling mistake `desnet` to `densenet` in densenet.md

* Update densenet.md

fix another mistake in this file

* add functions for images augmetation in image-augmentation.md

* add function img_aug to generate augmetation images in gluonbook/utils.py

* change a wrong word in neural-style.md