Clear Filters
Clear Filters

How ANN model deals with missing input data?

8 views (last 30 days)
I am new to matlab ANN, It will be very helpfull if somone can calrify these queries :
1) How ANN model deals with missing input data?
2) what will happen if I use my 100% data in training model ?
3) How to find out suitable no of layers
4) Why ANN model gives different results (R value) if I am training it again and again with the same input data
5) Can I input my data in sd form such as 185 ± 6
6) How to do process optimizatin using ANN model
Walter Roberson
Walter Roberson on 24 Nov 2023
2: If you use 100% of your data in training mode, you would typically end up with indexing errors or array-size-mismatch errors when it tried to construct the data to use in the test phase.
4: ANN training usually initializes the weights randomly. The whole process of ANN training involves generating a random set of initial weights, iterating to find better weights, recording the outcomes -- and then trying again and again... and eventually take the version that led to the best outcome.
5: No, data cannot be entered in the form of a ±
Sunita on 24 Nov 2023
Thank you so much for the quick reply

Sign in to comment.

Accepted Answer

Pratyush on 24 Nov 2023
Hi Sunita,
I understand that you have a couple of queries on MATLAB Artificial Neural Network. Here are the answers in sequential order to your queries:
  1. Missing data can be handled by imputing values using mean, median, or advanced methods like KNN imputation.
  2. Using all data for training can lead to overfitting. Split your data into training and testing sets for better generalization.
  3. Experiment with different numbers of layers, starting simple and gradually increasing complexity.
  4. Results can vary due to random weight initialization. Set a specific random seed for reproducibility.
  5. Yes, you can input data in SD form, but preprocess it to a format suitable for the neural network, such as normalization.
  6. Train the network to minimize or maximize an objective function using techniques like backpropagation, genetic algorithms, or particle swarm optimization in MATLAB.
Hope this helps.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!