I am looking for a machine learning, deep learning, deep reinforcement learning or any other classifier.
I will provide the data to you in .npy format (numpy binary format). The data is already separated in x and y data. The data is already in matrix (x) and columb (y). Both are validated. So it is not necessary to work to improve or format the data.
You may split the x and y in 2: training and test. I would suggest to split into 2 pieces not randomly picked, because each row in the x has information from previous rows. After the separation, you can shuffle the training data if you want.
I already have a model running and i have good results from it. The challenge is that you must provide a better solution. To validate that, another matrix with new x values will also be provided from my end. But in this case no y will be provided from my end. You shall provide the y from the matrix running your model/classifier and i will compare your results against mine. If you have a better results, i will buy and pay your solution otherwise i will not buy either pat for it.
No information about the data (what feature is about) will be provided. You can only eliminate a feature or more with my authorization to do it. Otherwise, you have to use all features.
You can use sklearn, tensor flow, rapids or openai.
In case you have a better solution than mine, the purchase will be performed and you shall provide the entire source code you use to train and execute the model. If your solution will NOT be better than mine, i will not pay you.
Even the training or the testing must able to be executed in a linux computer with 64GByte of Ram and 5Gbyte nvidia gpu card.
You can eventually change the hardware requirements, but you need my authorization to do so.
After the model/classifier is trained, the estimation for each row must be lower than 100ms using only one cpu.
You can use a voting classifier in case you use more than one classifier.
In case you want to use reinforcement or deep reinforcement learning, i can provide more information in how to implement the rewards.
You are welcome to ask any question about this work, but no information about what the data is about will be provided.
The delivery time is negotiable.