I have a simple Convolution1D model, that I have trained successfully
model = Sequential()
model.add(Embedding(input_dim=vocabsize, output_dim=32,
input_length=STR_MAX_LEN, dropout=0.2))
model.add(Dropout(0.2))
model.add(Convolution1D(64, 5, activation='relu', border_mode='same'))
model.add(Dropout(0.2))
model.add(MaxPooling1D())
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dropout(0.7))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss="binary_crossentropy", optimizer=Adam(), metrics=['accuracy'])
model.summary()
Model Summary as below
____________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
====================================================================================================
embedding_1 (Embedding) (None, 500, 32) 160000 embedding_input_1[0][0]
____________________________________________________________________________________________________
dropout_1 (Dropout) (None, 500, 32) 0 embedding_1[0][0]
____________________________________________________________________________________________________
convolution1d_1 (Convolution1D) (None, 500, 64) 10304 dropout_1[0][0]
____________________________________________________________________________________________________
dropout_2 (Dropout) (None, 500, 64) 0 convolution1d_1[0][0]
____________________________________________________________________________________________________
maxpooling1d_1 (MaxPooling1D) (None, 250, 64) 0 dropout_2[0][0]
____________________________________________________________________________________________________
flatten_1 (Flatten) (None, 16000) 0 maxpooling1d_1[0][0]
____________________________________________________________________________________________________
dense_1 (Dense) (None, 100) 1600100 flatten_1[0][0]
____________________________________________________________________________________________________
dropout_3 (Dropout) (None, 100) 0 dense_1[0][0]
____________________________________________________________________________________________________
dense_2 (Dense) (None, 1) 101 dropout_3[0][0]
====================================================================================================
Total params: 1770505
____________________________________________________________________________________________________
And I have a text that I need to run prediction on.
text = "dont know what could have saved limp dispiriting yam but it definitely wasnt a lukewarm mushroom as murky and appealing as bong water"
textWordsArray = np.array(text.split())
textIdxArrayPadded =
sequence.pad_sequences(textWordsIdxArray,maxlen=STR_MAX_LEN, value=0)
textIdxArrayPadded
structure of the text input
array([[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 5363, 121, 48, 97,
25, 1891, 8849, 51645, 19831, 18, 9, 404, 15422, 3, 15610, 27479, 14,
7217, 2, 2273, 14, 36597, 1090]], dtype=int32)
However, I am getting the below error when i run the prediction.
prediction = model.predict(textIdxArrayPadded, batch_size=1,verbose=1)
---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
<ipython-input-70-818365da75ca> in <module>()
----> 1 prediction = model.predict(textIdxArrayPadded, batch_size=1,verbose=1)
/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/models.pyc in predict(self, x, batch_size, verbose)
669 if self.model is None:
670 self.build()
--> 671 return self.model.predict(x, batch_size=batch_size, verbose=verbose)
672
673 def predict_on_batch(self, x):
/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/engine/training.pyc in predict(self, x, batch_size, verbose)
1177 f = self.predict_function
1178 return self._predict_loop(f, ins,
-> 1179 batch_size=batch_size, verbose=verbose)
1180
1181 def train_on_batch(self, x, y,
/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/engine/training.pyc in _predict_loop(self, f, ins, batch_size, verbose)
876 ins_batch = slice_X(ins, batch_ids)
877
--> 878 batch_outs = f(ins_batch)
879 if type(batch_outs) != list:
880 batch_outs = [batch_outs]
/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/backend/theano_backend.pyc in __call__(self, inputs)
715 def __call__(self, inputs):
716 assert type(inputs) in {list, tuple}
--> 717 return self.function(*inputs)
718
719
/home/ubuntu/anaconda2/lib/python2.7/site-packages/theano/compile/function_module.pyc in __call__(self, *args, **kwargs)
869 node=self.fn.nodes[self.fn.position_of_error],
870 thunk=thunk,
--> 871 storage_map=getattr(self.fn, 'storage_map', None))
872 else:
873 # old-style linkers raise their own exceptions
/home/ubuntu/anaconda2/lib/python2.7/site-packages/theano/gof/link.pyc in raise_with_op(node, thunk, exc_info, storage_map)
312 # extra long error message in that case.
313 pass
--> 314 reraise(exc_type, exc_value, exc_trace)
315
316
/home/ubuntu/anaconda2/lib/python2.7/site-packages/theano/compile/function_module.pyc in __call__(self, *args, **kwargs)
857 t0_fn = time.time()
858 try:
--> 859 outputs = self.fn()
860 except Exception:
861 if hasattr(self.fn, 'position_of_error'):
IndexError: One of the index value is out of bound. Error code: 65535.\n
Apply node that caused the error: GpuAdvancedSubtensor1(GpuElemwise{Composite{Switch(i0, (i1 * i2 * i3), i2)},no_inplace}.0, Elemwise{Cast{int64}}.0)
Toposort index: 38
Inputs types: [CudaNdarrayType(float32, matrix), TensorType(int64, vector)]
Inputs shapes: [(5000, 32), (500,)]
Inputs strides: [(32, 1), (8,)]
Inputs values: ['not shown', 'not shown']
Outputs clients: [[GpuReshape{3}(GpuAdvancedSubtensor1.0, MakeVector{dtype='int64'}.0)]]
HINT: Re-running with most Theano optimization disabled could give you a back-trace of when this node was created. This can be done with by setting the Theano flag 'optimizer=fast_compile'. If that does not work, Theano optimizations can be disabled with 'optimizer=None'.
HINT: Use the Theano flag 'exception_verbosity=high' for a debugprint and storage map footprint of this apply node.
I had the embeddings limited to the vocabsize, however I forgot to limit the word id to vocabsize
This was answered for me in a different forum , posting the solution here from the author.
#niazangels Niyas Mohammed Looks like you forgot to limit the
vocabulary to 5000 when encoding your test input!
Limit the vocabulary size to 5000
textWordsIdxArray = [np.array([i if i < vocabsize -1 else vocabsize -1 for i in s]) for s in textWordsIdxArray]
Related
I'm trying to train a model in Keras to suggest the best possible next move when presented with a pawn chess board. the board is represented as a list of 64 integers (0 for empty, 1 for player, 2 for enemy). The output is represented by a list of a field and a direction that the figure on that field should move in, which means I need two ouput layers with size 64 (number of fields) and 5 (number of possible move directions, including two forward and no move for when the game is over).
I have a list of boards and a list of solutions. When I try to fit the model however, I get the above mentioned error.
The exact error message is:
Epoch 1/75
Traceback (most recent call last):
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\main.py", line 75, in <module>
model.fit(train_fig_starts, train_fig_moves, epochs=75)
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\utils\traceback_utils.py", line 70, in error_handler
raise e.with_traceback(filtered_tb) from None
File "C:\Users\lulll\AppData\Local\Temp\__autograph_generated_filej0zia4d5.py", line 15, in tf__train_function
retval_ = ag__.converted_call(ag__.ld(step_function), (ag__.ld(self), ag__.ld(iterator)), None, fscope)
ValueError: in user code:
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1249, in train_function *
return step_function(self, iterator)
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1233, in step_function **
outputs = model.distribute_strategy.run(run_step, args=(data,))
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1222, in run_step **
outputs = model.train_step(data)
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1024, in train_step
loss = self.compute_loss(x, y, y_pred, sample_weight)
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\training.py", line 1082, in compute_loss
return self.compiled_loss(
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\engine\compile_utils.py", line 265, in __call__
loss_value = loss_obj(y_t, y_p, sample_weight=sw)
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\losses.py", line 152, in __call__
losses = call_fn(y_true, y_pred)
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\losses.py", line 284, in call **
return ag_fn(y_true, y_pred, **self._fn_kwargs)
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\losses.py", line 2176, in binary_crossentropy
backend.binary_crossentropy(y_true, y_pred, from_logits=from_logits),
File "C:\Users\lulll\Documents\CodeStuff\tfTesting\venv\lib\site-packages\keras\backend.py", line 5688, in binary_crossentropy
bce = target * tf.math.log(output + epsilon())
ValueError: Dimensions must be equal, but are 2 and 64 for '{{node binary_crossentropy/mul}} = Mul[T=DT_FLOAT](binary_crossentropy/Cast, binary_crossentropy/Log)' with input shapes: [?,2], [?,64].
I have absolutely no idea what is causing this. I've searched for the error already, but the only mentions I've found seem to be describing a completely different scenario.
Since it probably helps, here's the code used to create and fit the model:
inputs = tf.keras.layers.Input(shape=64)
x = tf.keras.layers.Dense(32, activation='relu')(inputs)
out_field = tf.keras.layers.Dense(64, name="field")(x)
out_movement = tf.keras.layers.Dense(5, name="movement")(x)
model = tf.keras.Model(inputs=inputs, outputs=[out_field, out_movement])
model.compile(optimizer='adam',
loss=tf.keras.losses.BinaryCrossentropy(),
metrics=['accuracy'])
model.fit(train_fig_starts, train_fig_moves, epochs=75) #train_fig_starts and moves are defined above
EDIT 1: Here's a sample of the dataset I'm using (the whole thing is too long for the character limit)
train_fig_starts = [[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 2, 2, 0, 1, 0, 0, 0, 0, 1, 2, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 2, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 2, 1, 0, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 2, 0, 0, 0, 0, 1], [0, 0, 1, 0, 0, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 2, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 2, 0], [0, 2, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 2, 1, 2, 2, 2, 0, 0, 0, 1, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 2, 2, 0, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 0, 1, 2, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0]]
train_fig_moves = [[0, 0], [0, 0], [0, 0], [0, 0], [15, 2], [15, 2]]
EDIT 2:
I changed it to sparsecategorialcrossentropy since that seems more like what I'm looking for. This is now the model code
inputs = tf.keras.layers.Input(shape=64)
x = tf.keras.layers.Dense(64, activation='relu')(inputs)
out_field = tf.keras.layers.Dense(64, activation="relu", name="field")(x)
out_field = tf.keras.layers.Dense(64, activation="softmax", name="field_softmax")(out_field)
out_movement = tf.keras.layers.Dense(5, activation="relu", name="movement")(x)
out_movement = tf.keras.layers.Dense(5, activation="softmax", name="movement_softmax")(out_movement)
model = tf.keras.Model(inputs=inputs, outputs=[out_field, out_movement])
print(model.summary())
tf.keras.utils.plot_model(model, to_file='model_plot.png', show_shapes=True, show_layer_names=True)
model.compile(optimizer='adam',
loss=[tf.keras.losses.SparseCategoricalCrossentropy(),
tf.keras.losses.SparseCategoricalCrossentropy()],
metrics=['accuracy'])
it still throws an error, this time its the following:
Node: 'sparse_categorical_crossentropy_1/SparseSoftmaxCrossEntropyWithLogits/SparseSoftmaxCrossEntropyWithLogits'
logits and labels must have the same first dimension, got logits shape [32,5] and labels shape [64]
[[{{node sparse_categorical_crossentropy_1/SparseSoftmaxCrossEntropyWithLogits/SparseSoftmaxCrossEntropyWithLogits}}]] [Op:__inference_train_function_1666]
I have no idea why its like that. Output logits and labels should both be [64, 2]. Since I'm using sparse crossentropy I should be able to use integers in my training data to signify the "index" of the ouput neuron with the highest logit, right? Correct me if I'm wrong. If it helps, here's a diagram of my model:
plot of the model
So I fixed the issue by myself now. Honestly it was a pretty stupid error to make but the error messages didn't really explain well what was going on. I swapped the outputs for one hot encoding and changed the loss to CategorialCrossEntropy, which is also more fitting for a categorisation problem (Sparse didn't work with my integers for some reason). After that I needed to change the label list from a 1dim list containing lists of len = 2 to a 2dim list containing both the field and the move one hots in a separate list. If anyone runs into a similar issue and can't make sense of it, maybe this will help.
I am using a FOR LOOP to calculate a simple probability on a dataset with approximately 500K rows of data.
For loop
class_ = 4
class_freq = Counter(list_[-1] for list_ in train_list) # Counter({5: 1476, 1: 1531, 4: 1562, 3: 1430, 2: 1498, 7: 1517, 6: 1486})
def cp(x, class_, freq_): # x is column index passed from another function
for row in train_list:
pos = 0
neg = 0
if row[x] == 1 and row[54] == class_:
pos+=1
else:
neg+=1
cal_0 = (neg + 0.1) / (class_freq[class_value] + 0.2)
cal_1 = (pos + 0.1) / (class_freq[class_value] + 0.2)
if prob_1 > prob_0:
return prob_1
else:
return prob_0
Train_list sample
[3050, 180, 4, 277, -3, 5782, 221, 242, 156, 2721, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1]
[2818, 119, 19, 30, 10, 5213, 248, 220, 92, 4497, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]
[3182, 115, 10, 553, 10, 4605, 237, 231, 124, 1768, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5]
[3024, 312, 18, 474, 177, 5785, 169, 224, 194, 4961, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]
[3067, 32, 4, 30, -2, 6679, 219, 230, 147, 2947, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4]
[2716, 1, 10, 234, 27, 2100, 206, 222, 153, 5581, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4]
...
The FOR LOOP works well on small dataset (few hundred rows) as expected. Unfortunately, when I try to use it on 20K rows of data, the processing time take ages. I cannot imagine how long it will take to run 500K rows of data.
FOR LOOP is grossly bad in performance for large dataset. What is an alternative to this? Will Lambda improve processing speed? I appreciate advice and assistance here, thanks.
Edited:
Thanks to everyone comments, I have tried to work on another algorithm to replace the FOR LOOP.
def cp(x, class_, class_):
filtered_list = [t for t in train_list if t[54] == class_]
count_binary = Counter(binary[col] for binary in filtered_list)
binary_1 = count_binary[1]
binary_0 = count_binary[0]
cal_0 = (binary_0 + 0.1) / (class_freq[class_value] + 0.2)
cal_1 = (binary_1 + 0.1) / (class_freq[class_value] + 0.2)
if prob_1 > prob_0:
return prob_1
else:
return prob_0
I am still running the above code in my program and the process is not done yet - so can't tell if it is much efficient. I will appreciate if someone can provide their opinion on this new block of code.
FYI, if this is indeed a better and more efficient code, then the issue of processing speed is most likely on other parts of my code.
Let say I have test.csv
filename
1 a.jpg
2 b.jpb
then I have test image folder
/test
test_dataset = torchvision.datasets.ImageFolder(root= path + 'test/',transform=trans)
this will bring all the test files
If I want to make a submission file after done training, how should I link test folder's name and submission.csv file name?
%%time
from torch.autograd import Variable
results = []
with torch.no_grad():
model.eval()
print('start')
for num, data in enumerate(test_loader):
#print(num)
imgs, label = data
imgs,labels = imgs.to(device), label.to(device)
test = Variable(imgs)
output = model(test)
ps = torch.exp(output)
top_p, top_class = ps.topk(1, dim = 1)
#print(top_class)
results += top_class.cpu().numpy().tolist()
predictions = np.array(results).flatten()
array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,
0,
1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,
0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])
How should I know which result is from which file?
Below I have the following code that generates a heatmap that plots each point as a block. But I want to switch the appearence to more traditional heatmaps. It currently looks like
but I want to make it appear like
though since the dataset is all 0 it would be one color but this is for future data. Below I have attached the code that generates the first heatmap, I need to rewrite the code to change its appearence into the second one. I couldnt find the code on the matplotlib examples
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
try:
temp = [
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0]]
temp = np.array(temp)
column = ["2-12","2-12","2-12","2-12", "2-13", "2-13","2-13","2-13","2-14","2-14","2-14","2-14", "2-15", "2-15", "2-15", "2-15", "2-16","2-16","2-16","2-16", "2-17", "2-17", "2-17", "2-17", "2-18","2-18","2-18","2-18","2-19","2-19","2-19","2-19", "2-20","2-20","2-20","2-20", "2-21", "2-21", "2-21", "2-21","2-22","2-22","2-22","2-22"]
nodes = ["0-3", "4-7", "8-11", "22-15", "26-19", "20-23", "24-27", "28-31", "32-35", "36-39"]
fig, ax = plt.subplots()
im = ax.imshow(temp)
# We want to show all ticks...
ax.set_xticks(np.arange(len(column)))
ax.set_yticks(np.arange(len(nodes)))
# ... and label them with the respective list entries
ax.set_xticklabels(column)
ax.set_yticklabels(nodes)
# Rotate the tick labels and set their alignment.
plt.setp(ax.get_xticklabels(), rotation=45, ha="right",
rotation_mode="anchor")
# Loop over data dimensions and create text annotations.
for i in range(len(nodes)):
for j in range(len(column)):
text = ax.text(j, i, temp[i, j],
ha="center", va="center", color="w")
fig.tight_layout()
plt.show()
except ValueError:
pass
Could you comment two version of variational autoencoder loss and show me why they give me different results?
Dataset:
data1 = np.array([0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0], dtype='int32')
data2 = np.array([1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0], dtype='int32')
data3 = np.array([0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0], dtype='int32')
100 samples each, so I have 300 samples.
Code 1:
def vae_loss(x, x_decoded_mean):
xent_loss = objectives.binary_crossentropy(x, x_decoded_mean)
kl_loss = -0.5 * K.mean(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var))
loss = xent_loss + kl_loss
return loss
vae.compile(optimizer='rmsprop', loss=vae_loss)
Code 2:
def zero_loss(y_true, y_pred):
return K.zeros_like(y_pred)
class CustomVariationalLayer(Layer):
def __init__(self, **kwargs):
self.is_placeholder = True
super(CustomVariationalLayer, self).__init__(**kwargs)
def vae_loss(self, x, x_decoded_mean):
xent_loss = original_dim * metrics.binary_crossentropy(x, x_decoded_mean)
K.exp(z_log_var), axis=-1)
return K.mean(xent_loss + kl_loss)
def call(self, inputs):
x = inputs[0]
x_decoded_mean = inputs[1]
loss = self.vae_loss(x, x_decoded_mean)
self.add_loss(loss, inputs=inputs)
return K.ones_like(x)
loss_layer = CustomVariationalLayer()([x, x_decoded_mean])
vae = Model(x, [loss_layer])
vae.compile(optimizer='rmsprop', loss=[zero_loss])
Results are so different and I don't see where? Latent dimension are different. Code 2 shows the separation between groups and code 1 not.
code 1, vae.predict... is not accurate and code 2 give me 1 on all features.
Code 2 gives me accurate feedback of the code:
sent_encoded = encoder.predict(np.array(test), batch_size = batch_size)
sent_decoded = generator.predict(sent_encoded)
and code 1 is not accurate at all.
Both experiments have the same layers. So, once again, where is the different and what is the best solution for dataset like described above?