How to increase the performance and accuracy of Deep neural network ?

I am a beginner in Neural network world. I want to make a seismic impedance inverse network. The code I used is given below. The accuracy and the performance is very low. What should I do to improve my network, any help will be extremely appreciable.

clc
clear
% Solve an Input-Output Fitting problem with a Neural Network
% Script generated by Neural Fitting app
% Created 14-May-2019 07:54:35
%
% This script assumes these variables are defined:
%
%   x - input data.
%   t - target data.

%Ricker_wave
f = 25;
fs = 150;
xval = -0.05:1/fs:0.05;
psi = (1-(2*(pi^2)*(f^2)*(xval.^2))).*(exp(-(pi^2)*(f^2).*(xval.^2)));

% % For visualise
% figure;
% plot(xval,psi)
% view([90 90])
% title('Mexican Hat Wavelet');
% grid on

%reflectivity
R1 = linspace(0,0.5,239580);

% For shuffling
[m1,l1] = size(R1) ;
idx1 = randperm(l1) ;
b1 = R1 ;
b1(1,idx1) = R1(1,:);

R2 = linspace(0,-1,239580);

[m2,l2] = size(R2) ;
idx2 = randperm(l2) ;
b2 = R2 ;
b2(1,idx2) = R2(1,:);

R3 = linspace(0.5,1,239580);

[m3,l3] = size(R3) ;
idx3 = randperm(l3) ;
b3 = R3 ;
b3(1,idx3) = R1(1,:);

RR = [R1;R2;R3];

% To make the Reflectivity in random places

RR1 = zeros(30,239580);

for j = 1:30
    T = randi(30,60,3);
    for k = 1:3
        n1 = round(T(k));
        if j==1
            RR1(n1,[1:7986]) = RR(k,[1:7986]);
        else
            RR1(n1,[((j-1)*7986)+1:j*7986]) = RR(k,[((j-1)*7986)+1:j*7986]);
    end
    end
end

% convolution

RR1000 = zeros(150,239580);
for i = 1:30
    if i == 1
        RR1000(30,:) = RR1(i,:);
    else
        RR1000(30+((i-1)),:) = RR1(i,:);
    end
end

ST = zeros(165,239580);

for i = 1:239580
    ST(:,i) = conv(RR1000(:,i),psi);
end

% To visualise the traces
% ST1 = ST(:,66200);
% figure;
% wiggle(ST1)
% grid on

x = ST(10:end,:);
t = RR1000;

% Choose a Training Function
% For a list of all training functions type: help nntrain

trainFcn = 'traingd';% Gradient-desent optimisation algorithm with momentum

% Create a Fitting Network
hiddenLayerSize = 200;
hiddenLayerSize1 = 100;
hiddenLayerSize2 = 50;
net = feedforwardnet([hiddenLayerSize hiddenLayerSize1...
    hiddenLayerSize2],trainFcn);

% maximum epoch as default is 1000
% performance goal is 0
% Minimum performance gradient is 1e-5
% Show training GUI is true
% Maximum time to train in seconds is inf

net.layers{1}.transferFcn = 'poslin';                     %Activation layer
net.trainParam.max_fail = 6;                %validation fail maximum number
net.trainParam.lr = 0.05;                                    %learning rate
net.performParam.regularization = 0.001;           %regularisation parametr
net.trainParam.mc = 0.1;                                   %moment constant

% Training stops when any of these conditions occurs:
% 
% The maximum number of epochs (repetitions) is reached.
% The maximum amount of time is exceeded.
% Performance is minimized to the goal.
% The performance gradient falls below min_grad.
% Validation performance has increased more than max_fail times since the 
% last time it decreased (when using validation).

% Setup Division of Data for Training, Validation, Testing
net.divideParam.trainRatio = 70/100;
net.divideParam.valRatio = 15/100;
net.divideParam.testRatio = 15/100;

% Train the Network
[net,tr] = train(net,x,t);

% Test the Network
y = net(x);
e = gsubtract(t,y);
performance = perform(net,t,y)

% View the Network
view(net)

% Plots
% Uncomment these lines to enable various plots.
%figure, plotperform(tr)
%figure, plottrainstate(tr)
%figure, ploterrhist(e)
%figure, plotregression(t,y)
%figure, plotfit(net,x,t)

%For testing data

y = net(ST(10:end,216200));

TT = 1:150;
TT = TT';

figure;
subplot(1,3,3)
stem(TT,y,'r')
view([90 90])
title('Computed reflectivity')
grid on

subplot(1,3,2)
stem(TT,RR1000(:,216200),'r')
view([90 90])
title('Actual reflectivity')
grid on

subplot(1,3,1)
ST1 = ST(10:end,216200);
plot(ST1)
view([90 90])
title('Trace')
grid on

title('Trace')
grid on