Dark Blue with Orange

1
Artificial Neural Networksfor Pattern Recognition
Jack Breese Computer Systems Quarter 4, Pd. 7
2
What is a Neural Network?

• Interconnected neurons
• Weights
• Output

3
Uses of Neural Networks

• Pattern Recognition
• Face Recognition
• OCR

4
Neurons

• Add up each weighted input
• Use an activation function to determine output
• Pass on output to next layer

5
Training Neural Networks

• Large input set
• Outputs are verified, weights adjusted along a
  gradient based on these results.For each neuron
  in the network For each connection to the
  neuron weight random_value() Until
  desired accuracy is reached For each example
  in the training data actual_out
  run_network(example) exp_out
  calculate_expected(example) error exp_out
  actual out For each neuron in the
  network calculate_delta_weights(error)
  adjust_weights(neuron)

6
Program Information

• Neural Network Library written in C
• Currently capable of initializing a two-layer
  perceptron with working, weighted connections.
• Capable of loading images and propagating data
  through the network.
• Can load images up to 500x500 pixels in size.

7
Data Structure
typedef struct _connection float
weight struct _neuron from
connection typedef struct _neuron //TODO
Implement a neuron which supports
connections. float d connection
cons neuron neuron mkneuron(int c)
neuron n malloc(sizeof(neuron)) n-gtd
0 connection a malloc(csizeof(connection))
n-gtcons a return n
8
New Progress

• Load PGM Images
• Create TrainingInfo structs
• Begin Training
• Perform Backpropagation

9
Training and Propagation Algos.
Calculating Neuron Values For each neuron in the
previous layer Sum neuron_weightneuron_val
ue neuron_value activation_function(sum)Train
ing For each neuron in the network For each
connection to the neuron weight
random_value() Until desired accuracy is
reached For each example in the training
data actual_out run_network(example)
exp_out calculate_expected(example) error
exp_out actual out For each neuron in
the network calculate_delta_weights(error)
adjust_weights(neuron)?
10
New Data Structures

• TrainInfo
• pImg

11
Testing

• Memory Usage was tested
• Training was attempted
• Values for known images and random weights
  propagated through.

12
Problems Encountered

• Initially thought memory usage was low.
• Forgot to reset counter in nested for loops to 0.
• That was dumb.
• Corrected problem, memory usage went up
• Decided to scale back network size/interconnectedn
  ess
• Issues with String arrays in C
• Prevented progress with training.

13
Conclusion

• Works as a valid header file
• Many methods
• Useful for further exploration