Paddle implementation, multidimensional time series data enhancement, mixup (using beta distribution to make continuous random numbers)

# Data to enhance 
def data_augment(X, y, p=0.8, alpha=0.5, beta=0.5):
    """Regression SMOTE
    1. Put the data x It is divided into  fix_X and X
    2. Yes X To remodel 
         For random uniform distribution less than 0.8 The index of idx_to_change Reshape the data part .
        X2 It's after disordering the order X
        beta The value is  np.random.beta(alpha, beta, batch_size) / 2 + 0.5 ####beta value >0.5  Make sure that most of the reshaped data is the original value 

        X[idx_to_change] = beta value *X[idx_to_change] +(1-beta value )*X2[idx_to_change]

    3. Merge fix_X and X New data x
    4. For input y In the same way 
    """
    fix_X, X = X[:, :, :, :2], X[:, :, :, 2:]#[32, 134, 144, 2] [32, 134, 144, 10]
    fix_y, y = y[:, :, :, :2], y[:, :, :, 2:]#[32, 134, 288, 2] [32, 134, 288, 10]
    # print('fix_X, X:',fix_X.shape, X.shape)
    batch_size = X.shape[0]#32
    random_values = paddle.rand([batch_size]) # Returns uniformly distributed , The scope is [0, 1) Of Tensor  shape:[32]
    idx_to_change = random_values < p #32 A value of , Less than 0.8 For the False, Others are True

    # ensure that first element to switch has probability > 0.5
    np_betas = np.random.beta(alpha, beta, batch_size) / 2 + 0.5
    random_betas = paddle.to_tensor(
        np_betas, dtype="float32").reshape([-1, 1, 1, 1]) # [32, 1, 1, 1]
    index_permute = paddle.randperm(batch_size)# returns a 1-D Tensor filled with random permutation values from 0 to n-1# Used to disrupt data numbers 

    X[idx_to_change] = random_betas[idx_to_change] * X[idx_to_change]
    X[idx_to_change] += (
        1 - random_betas[idx_to_change]) * X[index_permute][idx_to_change]

    y[idx_to_change] = random_betas[idx_to_change] * y[idx_to_change]
    y[idx_to_change] += (
        1 - random_betas[idx_to_change]) * y[index_permute][idx_to_change]
    return paddle.concat([fix_X, X], -1), paddle.concat([fix_y, y], -1)