Bug root cause : Wrong loss function calculation Bug symptom : Incorrect functionality Bug description : Because of using manually implemented cross-entropy loss, training loss shows nan value after a number of training iterations. So, they changed it to built-in softmax_cross_entropy_with_logits.
002
3.6
Correctness
Suboptimal learning rate
Bug root cause : suboptimal learning rate Bug symptom : Bad performance Bug description : Because of the fact that the learning rate is set to a value which is not small enough, it may lead to parameters blown up in the training process.
003
3.6
Correctness
Wrong loss function calculation
Bug root cause : Wrong loss function calculation Bug symptom : Incorrect functionality Bug description : Because of implementing cross_entropy manually, in some cases it shows nan value. Thus, they modified loss function calculation to avoid NAN value of training loss while model training.
004
3.6
Correctness
Wrong loss function calculation
Bug root cause : Wrong loss function calculation Bug symptom : Bad performance Bug description : Because of inaccuracy in the manually implemented loss functions, developers used softmax function in the right place to address this problem.
005
3.6
Correctness
Missing API call
Bug root cause : Missing API call Bug symptom : Bad performance Bug description : Because of problematic calculations in the manually implemented softmax layer, accuracy of the model stays near 10 %. Thus, they addressed this issue by porting the output of softmax function into built-in softmax layer type.
006
3.6
Correctness
Wrong Preprocessing
Bug root cause : Wrong preprocessing of train and test data Bug symptom : Crash Bug description : Because of wrong preprocessing of train and test data, training steps are faced with crash or nan value of cost in some cases.
007
3.6
Correctness
Wrong shape of input data
Bug root cause : Wrong shape of input data Bug symptom : Crash Bug description : Because of setting the wrong shape for input data, system faces with crash.
008
3.6
Correctness
Tensor shape mismatch
Bug root cause : Tensor shape mismatch Bug symptom : Crash Bug description : Because of a mismatched tensor shape in the network structure, system faces a crash.
009
3.6
Correctness
Wrong type of input data
Bug root cause : Wrong type of input data Bug symptom : Crash Bug description : Because of passing input data by wrong type, a compile error is raised.
010
3.6
Correctness
Wrong network architecture
Bug root cause : Wrong network architecture Bug symptom : Bad performance Bug description : Because of wrong model architecture, accuracy of the model is less than what is expected.
011
3.8
Model relevance
Suboptimal number of epochs
Bug root cause : Suboptimal number of epochs Bug symptom : Bad performance Bug description : Because of setting number of epochs to a value which is not optimal, accuracy of model will be lower than what is expected.
012
3.6
Correctness
Wrong network architecture
Bug root cause : Wrong network architecture Bug symptom : Bad performance Bug description : Because of bug in transformed channels, training step takes more than usual.
013
3.6
Correctness
Wrong weights initialisation
Bug root cause : Wrong weights initialisation Bug symptom : Crash Bug description : Because of the wrong place of loading checking, weights are initialized randomly.
014
3.6
Correctness
Deprecated API
Bug root cause : Deprecated API Bug symptom : Crash Bug description : Because of using deprecated API, system faces with crash.
015
3.7
Correctness
Wrong loss function calculation
Bug root cause : Wrong loss function calculation Bug symptom : Crash Bug description : Because of buggy loss function calculation, system faces with crash.
016
3.7
Correctness
Wrong network architecture
Bug root cause : Wrong network architecture Bug symptom : Crash Bug description : Because of wrong network architecture (Problem in residual blocks), systems faces with crash.
017
3.7
Correctness
Wrong loss function calculation
Bug root cause : Wrong loss function calculation Bug symptom : Bad performance Bug description : Because of buggy calculation of implemented loss function, accuracy of model is lower that what is expected.
018
3.7
Correctness
Deprecated API
Bug root cause : Deprecated API Bug symptom : Crash Bug description : Because of using deprecated APIs, system faces with crash.
019
3.7
Efficiency
Problematic early stopping
Bug root cause : Problematic early stopping Bug symptom : Bad performance Bug description : Because of wrong value of patience for early stopping, training takes longer than what is expected.
020
3.7
Correctness
Wrong network architecture
Bug root cause : Wrong network architecture Bug symptom : Crash Bug description : Because of wrong network architecture (Problem in residual blocks), systems faces with crash.
021
3.7
Correctness
Wrong selection of loss function
Bug root cause : Wrong selection of loss function Bug symptom : Bad performance Bug description : Because of choosing wrong loss function, accuracy of model stays at a lower level than what is should be.
022
3.7
Correctness
Suboptimal number of epochs
Bug root cause : Suboptimal number of epochs Bug symptom : Bad performance Bug description : Because of setting wrong value for epoch, accuracy of model stays at a lower level than what is should be.
023
3.8
Model relevance
Suboptimal number of epochs
Bug root cause : Suboptimal number of epochs Bug symptom : Bad performance Bug description : Because of setting wrong value for epoch, accuracy of model stays at a lower level than what is should be
024
3.8
Correctness
Wrong optimisation function
Bug root cause : Wrong optimisation function Bug symptom : Bad performance Bug description : Because of selecting wrong type of optimizer, accuracy of model stays at a lower level than what is should be.
025
3.6
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of suboptimal structure of the network, accuracy of the model stays at a value which is lower than what it should be.
026
3.6
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of choosing wrong type of activation function, accuracy of the model is less than what it should be.
027
3.6
Correctness
Missing dense layer
Bug root cause : Missing dense layer Bug symptom : Bad performance Bug description : Because of missing necessary dense layers, accuracy of the model stays at a lower level than what is expected
028
3.7
Correctness
Wrong API usage
Bug root cause : Wrong API usage Bug symptom : Crash Bug description : Because of wrong usage of Keras APIs, systems faces with crash.
029
3.8
Correctness
Deprecated API
Bug root cause : Deprecated API Bug symptom : Crash Bug description : Because of using deprecated APIs, system faces with crash
030
3.6
Correctness
Suboptimal batch size
Bug root cause : Suboptimal batch size Bug symptom : Bad performance Bug description : Because of setting wrong value as batch, accuracy of model stays at a lower level than what is should be.
031
3.7
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performanace Bug description : Because of designing network structure in suboptimal shape, model accuracy stays in lower level than what it should be.
032
3.6
Correctness
Wrong preprocessing
Bug root cause : Wrong preprocessing Bug symptom : Bad performance Bug description : Because of wrong place of upsampling, accuracy of the model is less than what is expected.
033
3.6
Correctness
Wrong selection of loss function
Bug root cause : Wrong selection of loss function Bug symptom : Bad performance Bug description : Because of selecting wrong loss function, accuracy of model stays at a lower level than what is should be.
034
3.6
Correctness
Missing preprocessing
Bug root cause : Missing preprocessing Bug symptom : Bad performance Bug description : Because of missing preprocessing (shuffle), accuracy of the model does not raise to the expected level.
035
3.7
Efficiency
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of designing network structure in suboptimal shape, speed of training is low.
036
3.7
Correctness
Wrong shape of input data
Bug root cause : Wrong shape of input data Bug symptom : Crash Bug description : Because of passing input data by wrong shape, a compile error is raised.
037
3.6
Correctness
Redundant data augmentation
Bug root cause : Redundant data augmentation Bug symptom : Bad performance Bug description : Because of buggy data augmentation, accuracy of the model stays at a lower level than what is should be.
038
3.7
Correctness
Suboptimal learning rate
Bug root cause : Suboptimal learning rate Bug symptom : Bad performance Bug description : Because of setting learning rate to a suboptimal value, model accuracy does not reach the level that is expected.
039
3.6
Correctness
Wrong API usage
Bug root cause : Wrong API usage Bug symptom : Crash Bug description : Because of using keras API in wrong way, a compile error is raised.
040
3.6
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of designing network structure in suboptimal shape, speed of training is low.
041
3.6
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of designing network structure in suboptimal shape, model accuracy does not reach the level that is expected.
042
3.6
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of designing network structure in suboptimal shape, model accuracy stays at a lower level than what is expected.
043
3.6
Correctness
Missing variable initialization
Bug root cause : Missing variable initialization Bug symptom : Crash Bug description : Because of problematic variable initialization, system faces with crash
044
3.6
Correctness
Epsilon for adam optimizer
Bug root cause : Epsilon for adam optimizer Bug symptom : Bad performance Bug description : Because of selecting a value as epsilon for adam optimizer which is not optimal, accuracy of the model is less than what is expected.
045
3.6
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of designing network structure in suboptimal shape, model accuracy stays at a lower level than what is expected.
046
3.6
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of using wrong type of activation function, model does not reach the expected accuracy level
047
3.6
Correctness
Suboptimal batch size
Bug root cause : Suboptimal batch size Bug symptom : Bad performance Bug description : Because of the suboptimal value of batch, model accuracy stays in a lower level than what it should be.
048
3.7
Correctness
Wrong filter size for convolution layer
Bug root cause : Wrong filter size for convolution layer Bug symptom : Bad performance Bug description : Because of setting wrong value for convolutional layer filter size, accuracy of the model is less than what is expected.
049
3.6
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of designing network structure in suboptimal shape, model accuracy stays at a lower level than what is expected.
050
3.6
Correctness
Suboptimal number of epochs
Bug root cause : Suboptimal number of epochs Bug symptom : Bad performance Bug description : Because of setting number of epochs to a value which is not optimal, accuracy of model will be lower than what is expected
051
3.6
Correctness
Missing dropout layer
Bug root cause : Missing dropout layer Bug symptom : Bad performance Bug description : Because of missing necessary dropout layer, accuracy of the model does not raise to the expected level.
052
3.6
Correctness
Suboptimal learning rate
Bug root cause : Suboptimal learning rate Bug symptom : Bad performance Bug description : Because of suboptimal learning rate, accuracy of the model is less than what is expected.
053
3.7
Correctness
Suboptimal number of epochs
Bug root cause : Suboptimal number of epochs Bug symptom : Bad performance Bug description : Because of suboptimal number of epochs, accuracy of the model is less than what is expected.
054
3.7
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Bad performance Bug description : Because of suboptimal network structure, accuracy of the model is less than what is expected.
055
3.7
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Crash Bug description : Because of designing model which done not have an optimal structure, memory overflow is occurred leading to system crash
056
3.7
Correctness
Wrong tensor shape
Bug root cause : Wrong tensor shape Bug symptom : Crash Bug description : Because of defining wrong shape of input tensor, systems faces with crash
057
3.6
Correctness
Deprecated API
Bug root cause : Deprecated API Bug symptom : Crash Bug description : Because of using deprecated APIs, system faces with crash
058
3.6
Correctness
Missing argument scoping
Bug root cause : Missing argument scoping Bug symptom : Crash Bug description : Because of ignorance of argument in calling APIs, a compile error is raised.
059
3.6
Correctness
Wrong optimisation function
Bug root cause : Wrong optimisation function Bug symptom : Crash Bug description : Because of selecting wrong optimizer function, system faces with crash.
060
3.6
Correctness
Wrong optimisation function
Bug root cause : Wrong optimisation function Bug symptom : Bad performance Bug description : Because of using wrong optimization function, accuracy of the model does not raise to the expected level.
061
3.8
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of selecting wrong activation function for output layer, model accuracy does not reach the expected accuracy level.
062
3.6
Correctness
Wrong layer type
Bug root cause : Wrong layer type Bug symptom : Bad performance Bug description : Because of using wrong layer type in the last layer, accuracy of the model stays at a lower level than what it should be
063
3.6
Correctness
Wrong filter size for convolutional layer
Bug root cause : Wrong filter size for convolutional layer Bug symptom : Bad performance Bug description : Because of setting wrong value for convolutional layer filter size, accuracy of the model is less than what is expected
064
3.6
Correctness
Deprecated API
Bug root cause : Deprecated API Bug symptom : Crash Bug description : Because of using deprecated Keras APIs, system faces with crash
065
3.6
Correctness
Wrong layer type
Bug root cause : Wrong layer type Bug symptom : Crash Bug description : Using wrong type of layer leads to system crash
066
3.8
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Incorrect functionality Bug description : Because of using the wrong activation function in last dense layer, the model could not predict negative values.
067
3.8
Correctness
Wrong selection of loss function
Bug root cause : Wrong selection of loss function Bug symptom : Bad performance Bug description : Because of using the wrong loss function, accuracy of the model does not reach the expected level.
068
3.8
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of using wrong activation function, accuracy of the model stays at a lower level than what is expected
069
3.8
Correctness
Missing preprocessing
Bug root cause : Missing preprocessing Bug symptom : Bad performance Bug description : Because of missing a preprocessing step (normalization), model does not reach the expected accuracy level.
070
3.8
Correctness
Suboptimal learning rate
Bug root cause : Suboptimal learning rate Bug symptom : Bad performance Bug description : Because of using default learning rate for SGD optimizer, model does not reach the best value of accuracy
071
3.8
Correctness
Wrong API usage
Bug root cause : Wrong API usage Bug symptom : Crash Bug description : Because of using mismatch API with Keras version, system faces with crash.
072
3.8
Correctness
Wrong selection of loss function
Bug root cause : Wrong selection of loss function Bug symptom : Bad performance Bug description : Because of using wrong loss function, accuracy of the model stays at a lower level than what is expected.
073
3.8
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Incorrect functionality Bug description : Because of selecting wrong activation function for output layer, model could not predict correctly.
074
3.6
Correctness
Wrong selection of loss function
Bug root cause : Wrong selection of loss function Bug symptom : Bad performance Bug description : Because of using wrong loss function, accuracy of the model is lower that what is expected.
075
3.6
Correctness
Suboptimal learning rate
Bug root cause : Suboptimal learning rate Bug symptom : Bad performance Bug description : Because of using default learning rate for adam optimizer, model does not reach the best value of accuracy.
076
3.7
Correctness
Deprecated API
Bug root cause : Deprecated API Bug symptom : Crash Bug description : Because of using deprecated Tensorflow APIs, system faced with compile error.
077
3.6
Correctness
Wrong type of input data
Bug root cause : Wrong type of input data Bug symptom : Crash Bug description : Passing the wrong type of input data to a method for saving Lambda layer, a compile error is raised.
078
3.8
Correctness
Suboptimal learning rate
Bug root cause : Suboptimal learning rate Bug symptom : Bad performance Bug description : Because of using default value of learning rate for optimizer which is suboptimal with respect to used data, accuracy of the model stays near 50 %.
079
3.6
Correctness
Wrong input format
Bug root cause : Wrong input format Bug symptom : Crash Bug description : Because of passing input data with wrong format, a compile error is raised.
080
3.6
Correctness
Missing Flatten layer
Bug root cause : Missing Flatten layer Bug symptom : Crash Bug description : Ignorance of adding necessary flatten layer to the model leads to compile error.
081
3.6
Correctness
Wrong shape of input data
Bug root cause : Wrong shape of input data Bug symptom : Crash Bug description : Because of passing input data with wrong shape, a compile error is raised.
082
3.7
Correctness
Wrong shape of input data
Bug root cause : Wrong shape of input data Bug symptom : Crash Bug description : Because of passing input data with wrong shape, a compile error is raised.
083
3.6
Correctness
Wrong loss function calculation
Bug root cause : Wrong loss function calculation Bug symptom : Crash Bug description : Because of a problem in the manually implemented loss functions, the system faces compile error.
084
3.6
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of using wrong activation function in dense layer, model does not reach the expected accuracy level
085
3.6
Correctness
Wrong shape of input data
Bug root cause : Wrong shape of input data Bug symptom : Crash Bug description : Because of passing input data with wrong shape, a compile error is raised.
086
3.7
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of using wrong activation function in dense layer, model does not reach the expected accuracy level.
087
3.7
Correctness
Wrong shape of input data
Bug root cause : Wrong shape of input data Bug symptom : Crash Bug description : Because of passing input data with wrong shape, a compile error is raised.
088
3.6
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of using wrong activation function of the last layer, model does not reach the expected accuracy level.
089
3.6
Correctness
Tensor shape mismatch
Bug root cause : Tensor shape mismatch Bug symptom : Crash Bug description : Because of passing data with mismatched tensor shape, a compile error is raised.
090
3.7
Correctness
Tensor shape mismatch
Bug root cause : Tensor shape mismatch Bug symptom : Crash Bug description : Because of passing tensors with mismatch shape between layers, a compile error raises.
091
3.7
Correctness
Missing preprocessing
Bug root cause : Missing preprocessing Bug symptom : Bad performance Bug description : Because of skipping necessary preprocessing step (data shuffling), accuracy of the model stays at near 0.
092
3.7
Correctness
Wrong type of activation function
Bug root cause : Wrong type of activation function Bug symptom : Bad performance Bug description : Because of using wrong activation function in dense layer, model does not reach the expected accuracy level.
093
3.7
Correctness
Tensor shape mismatch
Bug root cause : Tensor shape mismatch Bug symptom : Crash Bug description : Because of passing tensors with mismatch shape between layers, a compile error is raised.
094
3.6
Correctness
Wrong tensor shape
Bug root cause : Wrong tensor shape Bug symptom : Crash Bug description : Because of passing data with wrong tensor shape, a compile error is raised.
095
3.8
Correctness
Wrong API usage
Bug root cause : Wrong API usage Bug symptom : Bad performance Bug description : Because of wrong usage of Model APIs, model accuracy is lower than what it should be.
096
3.6
Correctness
Wrong API usage
Bug root cause : Wrong API usage Bug symptom : Crash Bug description : Because of wrong usage of fit API, a compile error is raised.
097
3.7
Correctness
Wrong input format
Bug root cause : Wrong input format Bug symptom : Crash Bug description : Because of wrong usage of confusion_matrix API, a compile error is raised.
098
3.8
Correctness
Wrong tensor shape
Bug root cause : Wrong tensor shape Bug symptom : Bad performance Bug description : Because of setting wrong tensor shape for test data, model does not reach the expected accuracy level.
099
3.8
Correctness
Suboptimal network structure
Bug root cause : Suboptimal network structure Bug symptom : Memory out of bound Bug description : Because of designing model which done not have an optimal structure, memory overflow is occurred leading to system crash.
100
3.7
Correctness
Missing preprocessing
Bug root cause : Missing preprocessing Bug symptom : Incorrect functionality Bug description : Because of missing a preprocessing step (data normalization), after some epochs nan value is shown as loss.