@@UnplacedIITian I didn't cover the installation in this video but you can find installation instructions for OpenAI baselines in their repository: github.com/openai/baselines
Thank you for the great work. More grace. Please do you think RL would be better suit for Intrusion Detection System (IDS) rather than supervised learning. or is it possible to have an hybrid of the two for IDS?
hi brthor, while training the MNIST classifier using DQN I Get an error "learn() got an unexpected keyword argument 'num_layers'" , do you have any idea how to resolve this?
Very good work ! As the RL classifier took more time than the supervised classifier, is it possible that it can be due to the fact that in the environnement (class MnistEnv), a single image is used rather than a batch ? and how to make it so that instead of a single image, we could take a batch of images in the environment? Furthermore, when using random=True in the class MnistEnv for the training, we are not sure that, each training image will be used at least 1 time, even if we specify 120,000 timesteps (2 times the number of samples in the training set). I would prefer random=False with a schuffle of the training data once the 60,000 images have been browsed. What do you think about it ?
Thanks for the video, Have you looked into the inference time? I was thinking maybe RL (PPO method) prediction time is faster than the supervised CNN model?
The inference is just a forward pass through the policy model in both cases, so the selected training algorithm will only change the architecture of that model. In this case all models have similar parameter numbers so I would expect inference times to be roughly equivalent.
Oh, Thank you 😊. The topic I was looking for. can you give me info about the requirements to install if I wanted to implement it in PyTorch? Thank you.
This is an issue with some mismatched versions. IIRC something with baselines and tensorflow, did you install dependencies using the included requirements.txt?
IIRC this is an issue with the version of some dependency. Ensure you are using the included requirements.txt file to install dependencies: `pip install -r requirements.txt`
Great video! I would like to try a Reinforcement Learning approach for a multi-label or binary classification of a CSV file. How would I set up my observation space and total timestep?
![[안방1열 풀캠4K] 베이비몬스터 'DRIP' (BABYMONSTER FullCam)│@SBS Inkigayo 241110](http://i.ytimg.com/vi/AQ4RIghEvUs/mqdefault.jpg)
I hope you learn something useful from this video.
Read the Blog Post: bit.ly/3vSOyu5
I am having problem importing baseline bcoz the module is no more available. Can you provide any alternative for that..
@@UnplacedIITian I didn't cover the installation in this video but you can find installation instructions for OpenAI baselines in their repository: github.com/openai/baselines
This is good stuff.
Thank you! You have answered to many of my questions in 7 minutes!
💎
what about reward functions?
Rewards is compulsory for classification or not?
great video! is it possible to re-upload the notebook, the old link isn't working for me
Thank you for the great work. More grace. Please do you think RL would be better suit for Intrusion Detection System (IDS) rather than supervised learning. or is it possible to have an hybrid of the two for IDS?
hi brthor,
while training the MNIST classifier using DQN I Get an error "learn() got an unexpected keyword argument 'num_layers'" , do you have any idea how to resolve this?
How to combined reinforcement based transfer learning
Really great video man! Keep up the good work!
Much appreciated 🙌
Very good work ! As the RL classifier took more time than the supervised classifier, is it possible that it can be due to the fact that in the environnement (class MnistEnv), a single image is used rather than a batch ? and how to make it so that instead of a single image, we could take a batch of images in the environment?
Furthermore, when using random=True in the class MnistEnv for the training, we are not sure that, each training image will be used at least 1 time, even if we specify 120,000 timesteps (2 times the number of samples in the training set). I would prefer random=False with a schuffle of the training data once the 60,000 images have been browsed. What do you think about it ?
can u provide stable-baselines for reinforcement learning classification? Unable to install baselines
try installing from the requirements file, it uses the github url for baselines IIRC
Hello, Could you please share me the updated link to requirements.txt file and notebook as the link is not active
Thanks for the video,
Have you looked into the inference time? I was thinking maybe RL (PPO method) prediction time is faster than the supervised CNN model?
The inference is just a forward pass through the policy model in both cases, so the selected training algorithm will only change the architecture of that model. In this case all models have similar parameter numbers so I would expect inference times to be roughly equivalent.
amazing work!!!!! like it !!!!! thank you so much
Oh, Thank you 😊. The topic I was looking for. can you give me info about the requirements to install if I wanted to implement it in PyTorch? Thank you.
hello brthor, please help how dealt with logger has no attribute confirgure
This is an issue with some mismatched versions. IIRC something with baselines and tensorflow, did you install dependencies using the included requirements.txt?
I am getting this error ModuleNotFoundError: No module named 'baselines' when I run the code.Please help.
There is a requirements.txt file included with the source. Run `pip install -r requirements.txt`
How could I predict the class for a single instance?
can it render env? how can i have GPU testing by bot?
In the training dqn part there is a error that says configure is not a attribute of logger. Can you please help me??
same problem with me
IIRC this is an issue with the version of some dependency. Ensure you are using the included requirements.txt file to install dependencies: `pip install -r requirements.txt`
Great video! I would like to try a Reinforcement Learning approach for a multi-label or binary classification of a CSV file. How would I set up my observation space and total timestep?
You'll need to adapt the code from the video to fit your data.
@@brthor1117 thank you! Would it make sense to use RL for predicting loan defaulters?
@@meryljacob6597 sounds like a better fit for supervised learning
@Meryl Jacob are you find solution ?
But repeated testing phase will improve the accuracy of RL model ,since it learns from the mistakes. Is it correct?
Training longer should improve the performance of the model.
Man you are awesome.
Link is not given
I updated the link in the description.
Can you provide link
I updated the link in the description.