Skip to content

ATTPC Noise Addition

AT-TPC Noise Addition

Only for use with simulated data

This method is used only when training with simulated data.

It is very difficult to exactly replicate the AT-TPC noise, but we can closely mimic it by creating noise that is uniformly random in z with np.random.randint(0,1000) and Gaussian in r with r_noise = np.random.normal(0, 50, (data_size,1)).

Can be used to add AT-TPC like noise by using np.random.randint() between [-50,50] for x and y, and [0,1000] for z.

Bases: BaseEstimator, TransformerMixin

Adding noise in cylindrical coordinates to simulate AT-TPC noise

Parameters

ratio_noise : (float) Fraction of noise to length of point cloud to be produced

Returns

new_data: (array) Data with AT-TPC noise added event_lengths: (array) New event lengths accounting for added noise points

Source code in scripts/ml_preprocessing_steps.py 
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
class AttpcNoiseAddition(BaseEstimator,TransformerMixin):
    """Adding noise in cylindrical coordinates to simulate AT-TPC noise

    Parameters
    ----------
    ratio_noise : (float)
        Fraction of noise to length of point cloud to be produced

    Returns
    ----------
    new_data: (array)
        Data with AT-TPC noise added
    event_lengths: (array)
        New event lengths accounting for added noise points 
    """
    def __init__(self, ratio_noise: float):
        self.ratio_noise = ratio_noise


    def fit(self,X,y=None):
        return self  

    def transform(self,X,y=None):
        """Adding AT-TPC like noise with cylindrical dataset

        Args:
            X (tuple):Packed data and event lengths np.array
            y (None): Defaults to None.

        Returns:
            (tuple): Data with noise and new event lengths
        """
        data,event_lengths = X
        skipped = 0
        for i in tqdm.tqdm(range(len(data)), desc="Adding Noise"): 
            data_size = int((self.ratio_noise)*event_lengths[i])

            noise_z = np.random.randint(0,1000,(data_size,1))
            r_noise = np.random.normal(0, 50, (data_size,1))

            theta_noise = np.random.uniform(0, 2*np.pi, (data_size,1))

            noise_x = r_noise * np.cos(theta_noise)
            noise_y = r_noise * np.sin(theta_noise)

            array_charge = np.zeros((data_size,1))
            noise_data = np.concatenate((noise_x,noise_y,noise_z,array_charge),axis=1)
            combined_data = np.concatenate((data[i,:event_lengths[i],:],noise_data,data[i,-2:,:]),axis=0)

            if combined_data.shape[0] > data.shape[1]:
                skipped+=1
                continue

            data[i, :combined_data.shape[0], :] = combined_data

            event_lengths[i]+=data_size


        print(f"Number of events skipped: {skipped}")
        return (data,event_lengths)

transform(X, y=None)

Adding AT-TPC like noise with cylindrical dataset

Parameters:

Name Type Description Default
X tuple

Packed data and event lengths np.array

 required
y None

Defaults to None.

 None

Returns:

Type Description
tuple

Data with noise and new event lengths
Source code in scripts/ml_preprocessing_steps.py 
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
def transform(self,X,y=None):
    """Adding AT-TPC like noise with cylindrical dataset

    Args:
        X (tuple):Packed data and event lengths np.array
        y (None): Defaults to None.

    Returns:
        (tuple): Data with noise and new event lengths
    """
    data,event_lengths = X
    skipped = 0
    for i in tqdm.tqdm(range(len(data)), desc="Adding Noise"): 
        data_size = int((self.ratio_noise)*event_lengths[i])

        noise_z = np.random.randint(0,1000,(data_size,1))
        r_noise = np.random.normal(0, 50, (data_size,1))

        theta_noise = np.random.uniform(0, 2*np.pi, (data_size,1))

        noise_x = r_noise * np.cos(theta_noise)
        noise_y = r_noise * np.sin(theta_noise)

        array_charge = np.zeros((data_size,1))
        noise_data = np.concatenate((noise_x,noise_y,noise_z,array_charge),axis=1)
        combined_data = np.concatenate((data[i,:event_lengths[i],:],noise_data,data[i,-2:,:]),axis=0)

        if combined_data.shape[0] > data.shape[1]:
            skipped+=1
            continue

        data[i, :combined_data.shape[0], :] = combined_data

        event_lengths[i]+=data_size


    print(f"Number of events skipped: {skipped}")
    return (data,event_lengths)