Time taken for prediction using competlayer Neural network.

I have an input of 2000 samples 800predictors, input is 2000x800,

using the following simple code

inputs = input';
net = competlayer(2);
net = train(net,inputs);
view(net)

outputs = net(input','useParallel','yes');
classes = vec2ind(outputs);
tabulate(classes)

calculating outputs is taking 2 seconds for each sample!

Can someone point out anything I am doing wrong. I want to reduce the time it takes to get output.

I know the number of predictors is really high but the etwork is giving good results so I dont want to change that.