深度神经网络可解释性：卷积核、权重和激活可视化（pytorch+tensorboard）

文章目录

引言
卷积核可视化
参数直方图可视化
激活可视化
小结

引言

一直以来，深度神经网络作为一种功能强大的“黑盒”，被认为可解释性较弱。目前，常用的一种典型可解释性分析方法是就是可视化方法。

本文整理了深度神经网络训练过程中常用的可视化技巧，便于对训练过程进行分析和检查。

卷积核可视化

以resnet18为例，提取第一层的卷积核（7x7）进行可视化，可以看出大多提取的是边缘、角点之类的底层视觉特征。

在全连接层前的卷积层采用的是3x3卷积核，表达高层语义信息，更加抽象：

这里采用torchvision.utils.make_grid对卷积核进行网格化显示，图像网格的列数由nrow参数确定。

卷积核的可视化代码参考¹进行修改：

def plot_conv(writer,model):for name,param in model.named_parameters():if 'conv' in name and 'weight' in name:in_channels = param.size()[1] # 输入通道out_channels = param.size()[0]   # 输出通道k_w, k_h = param.size()[3], param.size()[2]   # 卷积核的尺寸kernel_all = param.view(-1, 1, k_w, k_h)  # 每个通道的卷积核kernel_grid = torchvision.utils.make_grid(kernel_all, normalize=True, scale_each=True, nrow=in_channels)writer.add_image(f'{name}_all', kernel_grid, global_step=0)

参数直方图可视化

利用直方图可以对每一层参数的分布进行直观展示，便于分析模型参数的学习情况。

全连接层的参数分布如下图所示：

代码示例如下：

def plot_param_hist(writer,model):for name, param in model.named_parameters():writer.add_histogram(f"{name}", param, 0)

激活可视化

输入图像经过第一个卷积层的激活映射：

经过layer2和layer3的激活：

从pytorch模型中获取指定层的权重和激活的代码如下，参考facebook的工程²：

class GetWeightAndActivation:"""A class used to get weights and activations from specified layers from a Pytorch model."""def __init__(self, model, layers):"""Args:model (nn.Module): the model containing layers to obtain weights and activations from.layers (list of strings): a list of layer names to obtain weights and activations from.Names are hierarchical, separated by /. For example, If a layer follow a path"s1" ---> "pathway0_stem" ---> "conv", the layer path is "s1/pathway0_stem/conv"."""self.model = modelself.hooks = {}self.layers_names = layers# eval modeself.model.eval()self._register_hooks()def _get_layer(self, layer_name):"""Return a layer (nn.Module Object) given a hierarchical layer name, separated by /.Args:layer_name (str): the name of the layer."""layer_ls = layer_name.split("/")prev_module = self.modelfor layer in layer_ls:prev_module = prev_module._modules[layer]return prev_moduledef _register_single_hook(self, layer_name):"""Register hook to a layer, given layer_name, to obtain activations.Args:layer_name (str): name of the layer."""def hook_fn(module, input, output):self.hooks[layer_name] = output.clone().detach()layer = get_layer(self.model, layer_name)layer.register_forward_hook(hook_fn)def _register_hooks(self):"""Register hooks to layers in `self.layers_names`."""for layer_name in self.layers_names:self._register_single_hook(layer_name)def get_activations(self, input, bboxes=None):"""Obtain all activations from layers that we register hooks for.Args:input (tensors, list of tensors): the model input.bboxes (Optional): Bouding boxes data that might be requiredby the model.Returns:activation_dict (Python dictionary): a dictionary of the pair{layer_name: list of activations}, where activations are outputs returnedby the layer."""input_clone = [inp.clone() for inp in input]if bboxes is not None:preds = self.model(input_clone, bboxes)else:preds = self.model(input_clone)activation_dict = {}for layer_name, hook in self.hooks.items():# list of activations for each instance.activation_dict[layer_name] = hookreturn activation_dict, predsdef get_weights(self):"""Returns weights from registered layers.Returns:weights (Python dictionary): a dictionary of the pair{layer_name: weight}, where weight is the weight tensor."""weights = {}for layer in self.layers_names:cur_layer = get_layer(self.model, layer)if hasattr(cur_layer, "weight"):weights[layer] = cur_layer.weight.clone().detach()else:logger.error("Layer {} does not have weight attribute.".format(layer))return weights

对给定输入进行测试，输出指定层的激活映射，并绘制在tensorboard中：

# 模型测试，避免改变权重
model.eval()# Set up writer for logging to Tensorboard format.
writer = tb.TensorboardWriter(cfg)# 注册指定层的激活hook
layer_ls=["conv1","layer1/1/conv2","layer2/1/conv2","layer3/1/conv2","layer4/1/conv2"]
model_vis = GetWeightAndActivation(model, layer_ls)# 给定一个输入，获取指定层的激活映射
activations, preds = model_vis.get_activations(inputs)# 绘制激活映射（如画在tensorboard中）
plot_weights_and_activations(writer,activations,tag="Input {}/Activations: ".format(0))

小结

本文整理了深度神经网络常用的局部可视化代码，对卷积核、权重和激活映射进行可视化，便于对训练过程进行分析和检查。有需要的朋友可以马住收藏。

https://zhuanlan.zhihu.com/p/54947519 ↩︎
https://github.com/facebookresearch/SlowFast ↩︎