I hear the terms stability/instability thrown around a lot when reading up on Deep Q Networks. I understand that stability is improved with the addition of a target network and replay buffer but I fail to understand exactly what it's refering to.
What would the loss graph look like for an instable vs stable neural network?
What does it mean when a neural network converges/diverges?
Stability, also known as algorithmic stability, is a notion in
computational learning theory of how a machine learning algorithm is
perturbed by small changes to its inputs. A stable learning algorithm
is one for which the prediction does not change much when the training
data is modified slightly.
Here Stability means suppose you have 1000 training data that you use to train the model and it performs well. So in terms of model stability if you train the same model with 900 training data the model should still perform well , thats why it is also called as algorithmic stability.
As For the loss Graph if the model is stable the loss graph probably should be same for both size of training data (1000 & 900). And different in case of unstable model.
As in Machine learning we want to minimize loss so when we say a model converges we mean to say that the model's loss value is within acceptable margin and the model is at that stage where no additional training would improve the model.
Divergence is a non-symmetric metrics which is used to measure the difference between continuous value. For example you want to calculate difference between 2 graphs you would use Divergence instead of traditional symmetric metrics like Distance.
By simulating a tff code with random choice of clients in each round, I find that the accuracy increases to 0.9 then relapses to 0.5 and then from 0.8 to 0.6 and so on, it is not increasing.
have you any idea?
Thank you!
By chance is this a comparison of training accuracy and evaluation accuracy? In federated learning these two have different interpretations than the standard centralized machine learning setting.
In federated learning, training accuracy is often hard to interpret because it is the accuracy over client local model on the client local dataset. The client local dataset is a subset of the global distribution, with potentially a very different distribution.
Especially if there is high amounts of client drift (taking many client local training steps on a client local dataset that has large distribution differences), after averaging the individual client updates, the progress of the global model maybe far less than what the individual clients each round. Hence the evaluation accuracy at the end of the round maybe quite a bit worse than the training accuracy. In a "healthy" training process, this is expected to catch up after enough rounds have occurred.
I'm running a SAC reinforcement learner for a robotics application with some pretty decent results. One of the reasons I opted for reinforcement learning is for the ability for learning in the field, e.g. to adjust to a mechanical change, such as worn tires or a wheel going a little out of alignment.
My reinforcement learner restores it's last saved weights and replay buffer upon startup, so it doesn't need to retrain every time I turn it on. However, one concern I have is with respect to the optimizer.
Optimizers have come a long way since ADAM, but everything I read and all the RL code samples I see still seem to use ADAM with a fixed learning rate. I'd like to take advantage of some of the advances in optimizers, e.g. one cycle AdamW. However, a one-cycle optimizer seems inappropriate for a continuous real-world reinforcement learning problem: I imagine it's pretty good for the initial training/calibration, but I expect the low final learning rate would react too slowly to mechanical changes.
One thought I had was perhaps to do a one-cycle approach for initial training, and triggering a smaller one-cycle restart if a change in error that indicates something has changed (perhaps the size of the restart could be based on the size of the change in error).
Has anyone experimented with optimizers other than ADAM for reinforcement learning or have any suggestions for dealing with this sort of problem?
Reinforcement learning is very different from traditional supervised learning because the training data distribution changes as the policy improves. In optimization terms, the objective function can be said to be non-stationary. For this reason, I suspect your intuition is likely correct -- that a "one-cycle" optimizer would perform poorly after a while in your application.
My question is, what is wrong with Adam? Typically, the choice of optimizer is a minor detail for deep reinforcement learning; other factors like the exploration policy, algorithmic hyperparameters, or network architecture tend to have a much greater impact on performance.
Nevertheless, if you really want to try other optimizers, you could experiment with RMSProp, Adadelta, or Nesterov Momentum. However, my guess is that you will see incremental improvements, if any. Perhaps searching for better hyperparameters to use with Adam would be a more effective use of time.
EDIT: In my original answer, I made the claim that the choice of a particular optimizer is not primarily important for reinforcement learning speed, and neither is generalization. I want to add some discussion that helps illustrate these points.
Consider how most deep policy gradient methods operate: they sample a trajectory of experience from the environment, estimate returns, and then conduct one or more gradient steps to improve the parameterized policy (e.g. a neural network). This process repeats until convergence (to a locally optimal policy).
Why must we continuously sample new experience from the environment? Because our current data can only provide a reasonable first-order approximation within a small trust region around the policy parameters that were used to collect that data. Hence, whenever we update the policy, we need to sample more data.
A good way to visualize this is to consider an MM algorithm. At each iteration, a surrogate objective is constructed based on the data we have now and then maximized. Each time, we will get closer to the true optimum, but the speed at which we approach it is determined only by the number of surrogates we construct -- not by the specific optimizer we use to maximize each surrogate. Adam might maximize each surrogate in fewer gradient steps than, say, RMSProp does, but this does not affect the learning speed of the agent (with respect to environment samples). It just reduces the number of minibatch updates you need to conduct.
SAC is a little more complicated than this, as it learns Q-values in an off-policy manner and conducts updates using experience replay, but the general idea holds. The best attainable policy is subject to whatever the current data in our replay memory are; regardless of the optimizer we use, we will need to sample roughly the same amount of data from the environment to converge to the optimal policy.
So, how do you make a faster (more sample-efficient) policy gradient method? You need to fundamentally change the RL algorithm itself. For example, PPO almost always learns faster than TRPO, because John Schulman and co-authors found a different and empirically better way to generate policy gradient steps.
Finally, notice that there is no notion of generalization here. We have an objective function that we want to optimize, and once we do optimize it, we have solved the task as well as we can. This is why I suspect that the "Adam-generalizes-worse-than-SGD" issue is actually irrelevant for RL.
My initial testing suggest the details of the optimizer and it's hyperparameters matter, at least for off-policy techniques. I haven't had the chance to experiment much with PPO or on-policy techniques, so I can't speak for those unfortunately.
To speak to #Brett_Daley's thoughtful response a bit: the optimizer is certainly one of the less important characteristics. The means of exploration, and the use of a good prioritized replay buffer are certainly critical factors, especially with respect to achieving good initial results. However, my testing seems to show that the optimizer becomes important for the fine-tuning.
The off-policy methods I have been using have been problematic with fine-grained stability. In other words, the RL finds the mostly correct solution, but never really hones in on the perfect solution (or if it does find it briefly, it drifts off). I suspect the optimizer is at least partly to blame.
I did a bit of testing and found that varying the ADAM learning rate has an obvious effect. Too high and both the actor and critic bounce around the minimum and never converge on the optimal policy. In my robotics application this looks like the RL consistently makes sub-optimal decisions, as though there's a bit of random exploration with every action that always misses the mark a little bit.
OTOH, a lower learning rate tends to get stuck in sub-optimal solutions and is unable to adapt to changes (e.g. slower motor response due to low battery).
I haven't yet run any tests of single-cycle schedule or AdamW for the learning rate, but I did a very basic test with a two stage learning rate adjustment for both Actor and Critic (starting with a high rate and dropping to a low rate) and the results were a clearly more precise solution that converged quickly during the high learning rate and then honed in better with the low-learning rate.
I imagine AdamW's better weight decay regularization may result in similarly better results for avoiding overfitting training batches contributing to missing the optimal solution.
Based on the improvement I saw, it's probably worth trying single-cycle methods and AdamW for the actor and critic networks for tuning the results. I still have some concerns for how the lower learning rate at the end of the cycle will adapt to changes in the environment, but a simple solution for that may be to monitor the loss and do a restart of the learning rate if it drifts too much. In any case, more testing seems warranted.
I am a beginner in the neuronal network field and I want to understand a certain statement. A friend said that a neuronal network gets slower after you fit a lot of data in.
Right now, I just did the coursera ML course from androw ng. There, I implemented backpropagation. I thought it just adaptes the model related to the expected output by using different types of calculations. Nevertheless, it was not like the history was used to adapt the model. Just the current state of the neurons were checked and their weight were adapted backwards in combination with regularisation.
Is my assumption correct or am I wrong? Are there some libraries that use history data that could result in a slowly adapting model after a certain amount of training?
I want to use a simple neuronal network for reinforcement learning and I want to get an idea if I need to reset my model if the target environment changes for some reason. Otherwise my model would be slower and slower in adaption after time.
Thanks for any links and explanations in advanced!
As you have said, neural networks adapt by modifying their weights during the backpropagation step. Modifying these weights will not be slower as the training goes on since the number of steps to modify these weights will always remain the same. The amount of steps needed to run an example through your model will also remain the same, therefore not slowing down your network according to the amount of examples you fed it during training.
However, you can decide to change your learning rate during your training (generally decreasing it as epochs go on). According to the way the learning rate of your model evolves, the weights will be modified in a different manner, generally resulting in a smaller difference each epoch.
I was trying to train an emotion recognition model on the fer2013 dataset using the architecture proposed in this paper
The paper uses different dataset than mine so I did some modifications on on the stride and filter size.
After a couple hours of training, accuracy on both training and test set suddenly drops.
After that the accuracy just stay around 0.1-0.2 for both set, never improve anymore.
Does anybody know about this phenomenon?
In any neural network training, if both accuracies i.e. training and validation improves at first and then starts decreasing, it is a sign that your network is failing to converge. More appropriately, your optimizer has started overshooting.
One most likely reason for this could be high learning rate. Reduce your learning rate and then check your example again. Also, in your linked paper, (at least in first glimpse), I couldn't see learning rate mentioned. Since your data is different from the paper's, same learning rate might not work as well.