Trained neural networks

Examples of trained neural networks yielding a near-optimal policy are provided in the zoo directory.

They can be used simply by passing the name of the model to visualize.py or evaluate.py.

For example, to visualize an episode using the neural network policy defined by the model zoo_model_1_2_2, use:

python3 visualize.py --input zoo_model_1_2_2

The list of available models and corresponding parameters is given below. They are named zoo_model_i_j_k where i, j, k are integers associated to N_DIMS, LAMBDA_OVER_DX, R_DT.

Trained neural networks and their parameters.

N_DIMS

LAMBDA_OVER_DX

R_DT

FC_LAYERS

FC_UNITS

model_name

1

2.0

0.000001

3

128

zoo_model_1_zerohit

1

1.0

2.0

3

64

zoo_model_1_1_2

1

2.0

2.0

3

128

zoo_model_1_2_2

1

3.0

2.0

3

192

zoo_model_1_3_2

1

4.0

2.0

3

256

zoo_model_1_4_2

1

5.0

2.0

3

256

zoo_model_1_5_2

1

10

2.0

3

512

zoo_model_1_10_2

1

2.0

1e-2

3

128

zoo_model_1_2_001

1

2.0

1e-1

3

128

zoo_model_1_2_01

1

2.0

0.5

3

128

zoo_model_1_2_05

1

2.0

1

3

128

zoo_model_1_2_1

1

2.0

2

3

128

zoo_model_1_2_2

1

2.0

4

3

128

zoo_model_1_2_4

1

2.0

6

3

128

zoo_model_1_2_6

1

2.0

10

3

128

zoo_model_1_2_10

1

2.0

20

3

128

zoo_model_1_2_20

1

2.0

50

3

128

zoo_model_1_2_50

1

2.0

100

3

128

zoo_model_1_2_100

2

1.5

0.000001

3

256

zoo_model_2_zerohit

2

1.0

2.0

3

512

zoo_model_2_1_2

2

2.0

2.0

3

512

zoo_model_2_2_2

2

3.0

2.0

3

1024

zoo_model_2_3_2

2

1.0

1e-2

3

512

zoo_model_2_1_001

2

1.0

1e-1

3

512

zoo_model_2_1_01

2

1.0

0.5

3

512

zoo_model_2_1_05

2

1.0

1

3

512

zoo_model_2_1_1

2

1.0

2

3

512

zoo_model_2_1_2

2

1.0

4

3

128

zoo_model_2_1_4

2

1.0

6

3

128

zoo_model_2_1_6

2

1.0

10

3

128

zoo_model_2_1_10

2

1.0

20

3

128

zoo_model_2_1_20

2

1.0

50

3

128

zoo_model_2_1_50

2

1.0

100

3

128

zoo_model_2_1_100