{
    "model_type": "cnn",
    "architecture": {
        "type": "sequential",
        "layers": [
            {
                "type": "data_augmentation"
            },
            {
                "type": "conv2d",
                "filters": 32,
                "kernel_size": 3,
                "activation": "relu"
            },
            {
                "type": "batch_normalization"
            },
            {
                "type": "max_pooling2d",
                "pool_size": 2
            },
            {
                "type": "conv2d",
                "filters": 64,
                "kernel_size": 3,
                "activation": "relu"
            },
            {
                "type": "batch_normalization"
            },
            {
                "type": "max_pooling2d",
                "pool_size": 2
            },
            {
                "type": "flatten"
            },
            {
                "type": "dense",
                "units": 256,
                "activation": "relu"
            },
            {
                "type": "batch_normalization"
            },
            {
                "type": "dropout",
                "rate": 0.3
            },
            {
                "type": "dense",
                "units": 128,
                "activation": "relu"
            },
            {
                "type": "batch_normalization"
            },
            {
                "type": "dropout",
                "rate": 0.2
            },
            {
                "type": "dense",
                "units": 10,
                "activation": "softmax"
            }
        ]
    },
    "training_config": {
        "optimizer": {
            "type": "adam",
            "learning_rate": 0.001,
            "beta_1": 0.9,
            "beta_2": 0.999
        },
        "loss": "sparse_categorical_crossentropy",
        "metrics": [
            "accuracy"
        ],
        "epochs": 20,
        "batch_size": 32,
        "validation_split": 0.2
    },
    "preprocessing": {
        "input_normalization": "divide_by_255",
        "input_shape": [
            28,
            28,
            1
        ]
    }
}