Build neural network (classifier) + documentation report.
Budget
422$
per month
Posted: 4 year ago
Opened
- Description
- Job offer is on very short notice with a very short deadline. Will pay extra for work being completed quickly. More information will be provided in document attached.
Below are the implementations that I need done for my job ASAP. 3 Datasets will be used, 2 of which i will specify and third is your preference.
Implementing linear and ReLu layers
For this task, you should implement the forward and backward passes for linear layers and
ReLU activations.
Implementing dropout
For this task, you should implement inverted dropout. Forward and backward passes should
be implemented here.
Implementing softmax classifier
For this task, you should implement softmax loss and gradients. Explain numerical issues
when calculating the softmax function.
Implementing fully-connected NN
For this task, you should implement a fully-connected NN with arbitrary number of hidden
layers, ReLu activation, softmax classification and optional dropout. This task is about
reusing your implementations from Task 1 to Task 3. In addition, you will add a L2
regularizer. Report the parameters used (update rule, learning rate, decay, epochs, batch size)
and include the plots in your report.
Optimization of hyper-parameters
For this task, you should optimize the hyper-parameters of a fully-connected NN with your
chosen emotion classification dataset. Here you should implement SGD with momentum.
You need to select a performance measure which will be used to compare the performance of
the different parameters on the validation set.
Hyper-parameter optimization steps:
1. Choose a proper network architecture as starting point. Define a momentum and a
learning rate.
2. Try a different stopping criterion or a different learning rate update schedule.
3. Optimize the learning rate (disable regularization). Report how you found a good initial
value for learning rate. Include the plot for training loss and report training and validation
classification error.
4. Apply dropout and see if there is any improvement in the validation performance.
5. Compare L2 performance with dropout.
6. Finally, optimize the topology of the network (the number of hidden layers and the
number of neurons in each hidden layer).
A report will also have to be produced.
The report should cover each of the tasks above (and any other element of your work
that you believe should be reported). Graphically illustration of your results is expected
(perhaps training/testing error curves, confusion matrices, algorithm outputs, etc), as well as
results. Following the above analytical process, make sure you answer the following questions:
• What is your dataset, problem domain?
• Is your model classification or regression?
• Did you have any missing data? If so, how did you cope it?
• Did you do apply techniques to understand your dataset?
• What models did you use?
• How did you encode the input variables?
• What are the criteria for selecting model accuracy evaluation tools?
• What were your outputs?
• Did you have any problems or difficulties working with the dataset?
You should present the results clearly and concisely and provide a discussion of the results,
with conclusions related to problem being addressed.
Skills:
illustration,architectural design,artificial neural networks,algorithm development,classification,documentation
- Category
Source: peopleperhour.com
Drag