I'm getting really high variability in both the accuracy and loss between each epoch, as high as 10%. It happens to my accuracy all the time, and my loss when I start adding in dropout. However I really need the dropout, any ideas on how to smooth it out?
It is hard to say anything concrete without knowing what you do. But because you mentioned that your dataset is very small: 500 samples, I say that your 10% performance jumps are not surprising. Still a few ideas:
definitely use a bigger dataset if you can. If it is not possible to collect a bigger dataset, try to augment whatever you have.
try a smaller dropout and see how it goes, try different regularizers (dropout is not the only option)
you data is small, you can afford to run more than 200 iterations
see how your model performs on the test set, it is possible that it just severely overfitted the data
Beside the fact that the data set is very small, during a training with a dropout regularization the loss function is not anymore well defined and I presume the accuracy is also biased. Therefore any tracked metric should be assessed without dropout. It seams that keras does not switch it off while calculating the accuracy during training.
Related
I am trying some 1, 2, and 3 layer LSTM networks to classify land cover of some selected pixels from a Landsat time-series spectral data. I tried different optimizers (as implemented in Keras) to see which of them is better, and generally found AMSGrad variant of ADAM doing a relatively better job in my case. However, one strange thing to me is that for the AMSGrad variant, the training and test accuracies start at a relatively high value from the first epoch (instead of increasing gradually) and it changes only slightly after that, as you see in the below graph.
Performance of ADAM optimizer with AMSGrad on
Performance of ADAM optimizer with AMSGrad off
I have not seen this behavior in any other optimizer. Does it show a problem in my experiment? What can be the explanation for this phenomenon?
Pay attention to the number of LSTM layers. They are notorious for easily overfitting the data. Try a smaller model initially(less number of layers), and gradually increase the number of units in a layer. If you notice poor results, then try adding another LSTM layer, but only after the previous step has been done.
As for the optimizers, I have to admit I have never used AMSGrad. However, the plot with regard to the accuracy does seem to be much better in case of the AMSGrad off. You can see that when you use AMSGrad the accuracy on the training set is much better than that on the test set, which a strong sign of overfitting.
Remember to keep things simple, experiment with simple models and generic optimizers.
I am planning to make small Tensor Flow image classification project, which is expected to run on machines with low processing power, and one of the concerns I was asked about was the time needed to train the model.
The project is still in the conception stage and no clear boundary is made.
But assuming that we will use Tensor flow for Python, with a simple Neural Network for say n images data set, is there a way to estimate or predict the time required to train the model before performing the training given the hardware in use?
I have asked one of my colleagues who works in NN and he said that maybe we could calculate the time needed by measuring the time for the first epoch and making an estimation how many epochs needed afterwards. Is this is a valid way? If yes then is it even possible to estimate the number of epochs needed? And either cases is there a way to calculate it before performing any training?
There is no definite way of finding the number of epochs to which the model converges. It is one of the hyperparameter.
Apart from the type of model you are training, convergence also depends on the distribution of data, and the optimizer you are using.
The rough estimate you can make by looking at the number of parameters you have in your model, check time for one epoch, and get a rough idea from "experience" on the number of epochs. BUT you always have to look at the training and validation loss curves to check for the convergence.
I'm attempting to train a tensorflow model based on the popular slim implementation of mobilenet_v2 and am observing behaviour I cannot explain related (I think) to batch normalization.
Problem Summary
Model performance in inference mode improves initially but starts producing trivial inferences (all near-zeros) after a long period. Good performance continues when run in training mode, even on the evaluation dataset. Evaluation performance is impacted by batch normalization decay/momentum rate... somehow.
More extensive implementation details below, but I'll probably lose most of you with the wall of text, so here are some pictures to get you interested.
The curves below are from a model which I tweaked the bn_decay parameter of while training.
0-370k: bn_decay=0.997 (default)
370k-670k: bn_decay=0.9
670k+: bn_decay=0.5
Loss for (orange) training (in training mode) and (blue) evaluation (in inference mode). Low is good.
Evaluation metric of model on evaluation dataset in inference mode. High is good.
I have attempted to produce a minimal example which demonstrates the issue - classification on MNIST - but have failed (i.e. classification works well and the problem I experience is not exhibited). My apologies for not being able to reduce things further.
Implementation Details
My problem is 2D pose estimation, targeting Gaussians centered at the joint locations. It is essentially the same as semantic segmentation, except rather than using a softmax_cross_entropy_with_logits(labels, logits) I use tf.losses.l2_loss(sigmoid(logits) - gaussian(label_2d_points)) (I use the term "logits" to describe unactivated output of my learned model, though this probably isn't the best term).
Inference Model
After preprocessing my inputs, my logits function is a scoped call to the base mobilenet_v2 followed by a single unactivated convolutional layer to make the number of filters appropriate.
from slim.nets.mobilenet import mobilenet_v2
def get_logtis(image):
with mobilenet_v2.training_scope(
is_training=is_training, bn_decay=bn_decay):
base, _ = mobilenet_v2.mobilenet(image, base_only=True)
logits = tf.layers.conv2d(base, n_joints, 1, 1)
return logits
Training Op
I have experimented with tf.contrib.slim.learning.create_train_op as well as a custom training op:
def get_train_op(optimizer, loss):
global_step = tf.train.get_or_create_global_step()
opt_op = optimizer.minimize(loss, global_step)
update_ops = set(tf.get_collection(tf.GraphKeys.UPDATE_OPS))
update_ops.add(opt_op)
return tf.group(*update_ops)
I'm using tf.train.AdamOptimizer with learning rate=1e-3.
Training Loop
I'm using the tf.estimator.Estimator API for training/evaluation.
Behaviour
Training initially goes well, with an expected sharp increase in performance. This is consistent with my expectations, as the final layer is rapidly trained to interpret the high-level features output by the pretrained base model.
However, after a long period (60k steps with batch_size 8, ~8 hours on a GTX-1070) my model begins to output near-zero values (~1e-11) when run in inference mode, i.e. is_training=False. The exact same model continues to improve when run in *training mode, i.e.is_training=True`, even on the valuation set. I have visually verified this is.
After some experimentation I changed the bn_decay (batch normalization decay/momentum rate) from the default 0.997 to 0.9 at ~370k steps (also tried 0.99, but that didn't make much of a difference) and observed an immdeiate improvement in accuracy. Visual inspection of the inference in inference mode showed clear peaks in the inferred values of order ~1e-1 in the expected places, consistent with the location of peaks from training mode (though values much lower). This is why the accuracy increases significantly, but the loss - while more volative - does not improve much.
These effects dropped off after more training and reverted to all zero inference.
I further dropped the bn_decay to 0.5 at step ~670k. This resulted in improvements to both loss and accuracy. I'll likely have to wait until tomorrow to see the long-term effect.
Loss and an evaluation metric plots given below. Note the evaluation metric is based on the argmax of the logits and high is good. Loss is based on the actual values, and low is good. Orange uses is_training=True on the training set, while blue uses is_training=False on the evaluation set. The loss of around 8 is consistent with all zero outputs.
Other notes
I have also experimented with turning off dropout (i.e. always running the dropout layers with is_training=False), and observed no difference.
I have experimented with all versions of tensorflow from 1.7 to 1.10. No difference.
I have trained models from the pretrained checkpoint using bn_decay=0.99 from the start. Same behaviour as using default bn_decay.
Other experiments with a batch size of 16 result in qualitatively identical behaviour (though I can't evaluate and train simultaneously due to memory constraints, hence quantitatively analysing on batch size of 8).
I have trained different models using the same loss and using tf.layers API and trained from scratch. They have worked fine.
Training from scratch (rather than using pretrained checkpoints) results in similar behaviour, though takes longer.
Summary/my thoughts:
I am confident this is not an overfitting/dataset problem. The model makes sensible inferences on the evaluation set when run with is_training=True, both in terms of location of peaks and magnitude.
I am confident this is not a problem with not running update ops. I haven't used slim before, but apart from the use of arg_scope it doesn't look too much different to the tf.layers API which I've used extensively. I can also inspect the moving average values and observe that they are changing as training progresses.
Chaning bn_decay values significantly effected the results temporarily. I accept that a value of 0.5 is absurdly low, but I'm running out of ideas.
I have tried swapping out slim.layers.conv2d layers for tf.layers.conv2d with momentum=0.997 (i.e. momentum consistent with default decay value) and behaviour was the same.
Minimal example using pretrained weights and Estimator framework worked for classification of MNIST without modification to bn_decay parameter.
I've looked through issues on both the tensorflow and models github repositories but haven't found much apart from this. I'm currently experimenting with a lower learning rate and a simpler optimizer (MomentumOptimizer), but that's more because I'm running out of ideas rather than because I think that's where the problem lies.
Possible Explanations
The best explanation I have is that my model parameters are rapidly cycling in a manner such that the moving statistics are unable to keep up with the batch statistics. I've never heard of such behaviour, and it doesn't explain why the model reverts to poor behaviour after more time, but it's the best explanation I have.
There may be a bug in the moving average code, but it has worked perfectly for me in every other case, including a simple classification task. I don't want to file an issue until I can produce a simpler example.
Anyway, I'm running out of ideas, the debug cycle is long, and I've already spent too much time on this. Happy to provide more details or run experiments on demand. Also happy to post more code, though I'm worried that'll scare more people off.
Thanks in advance.
Both lowering the learning rate to 1e-4 with Adam and using Momentum optimizer (with learning_rate=1e-3 and momentum=0.9) resolved this issue. I also found this post which suggests the problem spans multiple frameworks and is an undocumented pathology of some networks due to the interaction between optimizer and batch-normalization. I do not believe it is a simple case of the optimizer failing to find a suitable minimum due to the learning rate being too high (otherwise performance in training mode would be poor).
I hope that helps others experiencing the same issue, but I'm a long way from satisfied. I'm definitely happy to hear other explanations.
I have a trained a neural network on a classification task, and it is learning, although it's accuracy is not high. I am trying to figure out which test examples it is not confident about, so that I can gain some more insight into what is happening.
In order to do this, I decided to use the standard softmax probabilities in Tensorflow. To do this, I called tf.nn.softmax(logits), and used the probabilities provided here. I noticed that many times the probabilities were 99%, but the prediction was still wrong. As such, even when I only consider examples who have prediction probabilities higher than 99%, I get a poor accuracy, only 2-3 percent higher than my original accuracy.
Does anyone have any ideas as to why the network is so confident about wrong predictions? I am still new to deep learning, so am looking for some ideas to help me out.
Also, is using the softmax probabilities the right way to do determine confidence of predictions from a neural network? If not, is there a better way?
Thanks!
Edit: From the answer below, it seems like my network is just performing poorly. Is there another way to identify which predictions the network makes are likely to be wrong besides looking at the confidence (since the confidence doesn't seem to work well)?
Imagine your samples are split by a vertical line but you NN classifier learnt a horizontal line, in this case any prediction given by your classifier can only obtain 50% accuracy always. However NN will assign higher confidence to the samples which are further away from the horizontal line.
In short, when your model is doing poor classification higher confidence has little to none contribution to accuracy.
Suggestion: Check if the information you needed to do the correct classification are in the data then improve the overall accuracy first.
This might be a weird question, buy here goes: I have a DNNClassifier in tensorflow that I ran with all sorts of parameters, and numbers of epochs. The weird thing that is happening is that, whatever I do, my accuracy on my train set stays quite low (It's a binary classifier, accuracy stays around 65%). I would expect that with at least some of the configurations I've tried (ones without regularization/dropout), I would've overfitted and reached a higher accuracy on my training set, but alas.
So, what does this mean? That I don't have enough data? That my network is not big enough to fit the data set? I tried doubling the size of my hidden layers, but it still stayed pretty much the same.
I have 23000 training examples, and currently have 3 layers of 100 neurons each.
Code can be found here: https://www.pastebucket.com/563290