【MXNet Symbol】 多指标评价模型性能

【承接图像分类、检测、分割、生成相关项目，私信。】

MXNet 分类模型训练之采用多指标评价模型能力（accuracy,cross-entropy,top_k_accuracy）

代码如下

metric=[mx.metric.create('acc'),
			mx.metric.create('top_k_accuracy', top_k=3),
			mx.metric.create('ce')]
    mod.fit(train, val,
        num_epoch=num_epoch,
        arg_params=arg_params,
        aux_params=aux_params,
        allow_missing=True,
        batch_end_callback = mx.callback.Speedometer(batch_size, 500),
        epoch_end_callback=checkpoint,
        kvstore='device',
        optimizer='sgd',
        optimizer_params={'learning_rate':0.00001,"momentum":0.9},
        initializer=mx.init.Xavier(rnd_type='gaussian', factor_type="in", magnitude=2),
        eval_metric=metric)
           

结果如下：

2017-09-06 16:46:25,648 Epoch[0] Batch [500]	Speed: 39.24 samples/sec	accuracy=0.929890	top_k_accuracy_3=0.983782	cross-entropy=0.264878
2017-09-06 16:49:49,687 Epoch[0] Batch [1000]	Speed: 39.21 samples/sec	accuracy=0.949125	top_k_accuracy_3=0.987875	cross-entropy=0.215848
2017-09-06 16:53:13,980 Epoch[0] Batch [1500]	Speed: 39.16 samples/sec	accuracy=0.960625	top_k_accuracy_3=0.991625	cross-entropy=0.179014