Forex Prediction Using Prototypical and Matching Networks
- or -
Post a project like this1422
$200
- Posted:
- Proposals: 2
- Remote
- #2838664
- Awarded
Description
Experience Level: Entry
Estimated project duration: less than 1 week
Urgent need to finish in 1 week maximum and must be able to run on python jupyter notebook with windows compatibility! (I do not want problems with mac/apple as i am a windows user)
Can use machine learning libraries like tensorflow, keras etc
I have a forex dataset for EUR/USD.
I need you to build a forex prediction model using few shot machine techniques such as Prototypical networks and Matching networks. I want to show that these methods are as effective to produce a high accuracy of near 95% at least in prediction and can be used instead of using large data. In both cases forex data must be converted to image data for few shot testing. Afterwhich, they must be tested again and tuned using these hyper parameters:
Hyperparameters for each network:
For Prototypical network,
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This all should be done in one notebook.
For Prototypical Network with Softmax FC ( network can be connected to a fully connected layer to enhance accuracy)
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This should be done in another notebook.
For Matching networks,
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This should be done in another notebook.
For Matching Network with any FCE ( network can be connected to a Full context embedding to enhance accuracy)
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This should be done in another notebook.
Overall, you have to produce 4 deliverables/notebooks that run:
1) Forex Prediction using Prototypical Networks
2) Forex Prediction using Prototypical Networks with Softmax FC
3) Forex Prediction using Matching Networks
4) Forex Prediction using Matching Network with any FCE
All notebooks must be given in maximum a week and i am no pro, so we can communicate everyday regarding progress.
Below I have attached word document of tables of the format output to give you an idea on what I need.
Also, below is some reference I have attached in file 'Fewshot2' showing a prototypical network notebook example and the input image data which you can use which are snippets and I use 25 images of each feature Close, Open, High, Low, Last. That will be your input data as well, which I can email you.
I have also attached a few links that might be helpful:
Prototypical networks - https://github.com/Hsankesara/Prototypical-Networks
Matching networks - https://github.com/AntreasAntoniou/MatchingNetworks
Thank you so much, really looking forward to working with you!
Can use machine learning libraries like tensorflow, keras etc
I have a forex dataset for EUR/USD.
I need you to build a forex prediction model using few shot machine techniques such as Prototypical networks and Matching networks. I want to show that these methods are as effective to produce a high accuracy of near 95% at least in prediction and can be used instead of using large data. In both cases forex data must be converted to image data for few shot testing. Afterwhich, they must be tested again and tuned using these hyper parameters:
Hyperparameters for each network:
For Prototypical network,
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This all should be done in one notebook.
For Prototypical Network with Softmax FC ( network can be connected to a fully connected layer to enhance accuracy)
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This should be done in another notebook.
For Matching networks,
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This should be done in another notebook.
For Matching Network with any FCE ( network can be connected to a Full context embedding to enhance accuracy)
1) Distance metric - Cosine, Euclidean, Euclidean^2
2) Number of episode - 5way 1shot testing OR 5way, 5shot testing
Calculate all log loss and average accuracy. This should be done in another notebook.
Overall, you have to produce 4 deliverables/notebooks that run:
1) Forex Prediction using Prototypical Networks
2) Forex Prediction using Prototypical Networks with Softmax FC
3) Forex Prediction using Matching Networks
4) Forex Prediction using Matching Network with any FCE
All notebooks must be given in maximum a week and i am no pro, so we can communicate everyday regarding progress.
Below I have attached word document of tables of the format output to give you an idea on what I need.
Also, below is some reference I have attached in file 'Fewshot2' showing a prototypical network notebook example and the input image data which you can use which are snippets and I use 25 images of each feature Close, Open, High, Low, Last. That will be your input data as well, which I can email you.
I have also attached a few links that might be helpful:
Prototypical networks - https://github.com/Hsankesara/Prototypical-Networks
Matching networks - https://github.com/AntreasAntoniou/MatchingNetworks
Thank you so much, really looking forward to working with you!
Faris A.
100% (4)Projects Completed
3
Freelancers worked with
3
Projects awarded
67%
Last project
30 May 2020
Singapore
New Proposal
Login to your account and send a proposal now to get this project.
Log inClarification Board Ask a Question
-
There are no clarification messages.
We collect cookies to enable the proper functioning and security of our website, and to enhance your experience. By clicking on 'Accept All Cookies', you consent to the use of these cookies. You can change your 'Cookies Settings' at any time. For more information, please read ourCookie Policy
Cookie Settings
Accept All Cookies