Skip to main content

Table 4 The number of successfully folding to the lowest energy conformations in the last 100 episodes

From: Research on predicting 2D-HP protein folding using reinforcement learning with full state space

Sequence No.

Methods

1

2

3

4

5

AVG

5

Full states

8

10

8

7

6

8

Greedy algorithm

7

7

8

5

7

7

Partial states

0

0

0

0

0

0

6

Full states

3

8

3

1

6

4

Greedy algorithm

9

11

6

3

10

8

Partial states

0

0

0

0

0

0

7

Full states

7

4

3

3

3

4

Greedy algorithm

7

5

7

9

9

7

Partial states

0

0

0

0

0

0

8

Full states

7

8

8

11

5

8

Greedy algorithm

2

3

2

6

2

3

Partial states

0

0

0

1

0

0

9

Full states

5

4

1

3

2

3

Greedy algorithm

0

0

0

1

1

0

Partial states

0

0

0

0

0

0

10

Full states

14

12

11

15

12

13

Greedy algorithm

12

9

9

17

8

11

Partial states

1

1

1

2

0

1

11

Full states

3

7

2

3

4

4

Greedy algorithm

4

0

0

2

1

1

Partial states

0

0

0

0

0

0

12

Full states

9

7

8

6

8

8

Greedy algorithm

2

5

2

5

2

3

Partial states

0

0

1

0

0

0

13

Full states

0

2

2

3

4

2

Greedy algorithm

0

0

0

0

0

0

Partial states

0

0

0

0

0

0

14

Full states

2

0

1

1

2

1

Greedy algorithm

0

0

0

0

0

0

Partial states

0

0

0

0

0

0

15

Full states

2

4

5

2

2

3

Greedy algorithm

0

0

0

0

0

0

Partial states

0

0

0

0

0

0

16

Full states

1

2

4

5

1

2

Greedy algorithm

1

0

0

0

0

0

Partial states

0

0

0

0

0

0

  1. The data in bold and italic indicates the average number of successfully folding to the lowest energy conformations by the reinforcement learning with full states is more than the other two methods