Archive of July 2012

it's amazing what changing the targets can do. now they're spread evenly from the starting position (uniform distance) and they seem to do much better, rather than finding a local maximum (a single target that's easier to get to than the others).

run length:27

	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25
avg	0.1299111	0.1520748	0.2072772	0.3187084	0.3958412	0.6468749	0.6010848	0.7655407	0.9283564	0.9560807	1.27025	1.240294	1.078598	1.257212	1.092487	1.327467	1.08835	1.154903	1.370362	1.198214	1.211951	1.178578	1.253744	0.9513875	1.347026	0.8872428
max	0.6187887	0.7656475	1.199693	1.060628	1.624553	2.014889	2.00272	1.532919	1.779993	2.924508	2.019435	1.890425	2.37094	2.055064	1.812547	2.125103	2.044051	1.91144	2.050521	2.57358	1.498748	1.417337	1.908807	1.261683	1.80807	1.39744

Population Size

100

Run Length

20

Do Training

False

Do Hybrid Training

True

Do Competitive Run

False

Do Adaptive Training

True

Inputs

hasTarget,targetLeft,targetRight,targetNorth,targetSouth,dirToTarget,distToTarget,wall,sensor0,sensor1,sensor2

Outputs

left,right,up,down,run,sensordir0,sensordir1,sensordir2

Hidden Layers

8 12 12 12 8

Back Propogation

True

Learning rate

0.5

Momentum

0.1

Growth rate

0.5

run length:48

	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31	32	33	34	35	36	37	38	39	40	41	42	43	44	45	46
avg	0.11333	0.1216922	0.1410408	0.2255169	0.3438947	0.5308622	0.7092866	0.7473413	0.8380993	0.8315367	0.9242428	0.8393147	1.103198	1.054069	1.111683	1.051629	1.028054	1.043636	0.9677835	1.01511	1.045051	1.138088	1.002189	0.8751988	0.9423944	0.9026732	1.062976	0.9172046	1.334726	1.079931	0.7032812	1.377699	1.236436	1.741362	1.081763	1.341366	0.8942567	0.9266959	1.079842	1.067705	0.7557345	1.354618	0.9338629	1.070021	1.288732	1.196118	1.037447
max	0.4842224	0.7318087	0.7872303	1.314361	1.167771	1.026197	1.422564	1.538746	1.474501	1.283453	1.423418	1.349116	1.614929	1.392465	1.4685	1.274314	1.203464	1.228827	1.08125	1.171278	1.106939	1.359823	1.106474	0.9218987	1.06358	1.007161	1.376667	0.9957669	1.447871	1.173347	0.8280649	1.515407	1.369468	1.839345	1.161362	1.377432	0.9441369	0.9500773	1.109659	1.092207	0.7708632	1.38819	0.9459927	1.093874	1.322104	1.24041	1.065003

Population Size

100

Run Length

20

Do Training

False

Do Hybrid Training

True

Do Competitive Run

False

Do Adaptive Training

True

Inputs

hasTarget,targetLeft,targetRight,targetNorth,targetSouth,dirToTarget,distToTarget,wall,sensor0,sensor1,sensor2

Outputs

left,right,up,down,run,sensordir0,sensordir1,sensordir2

Hidden Layers

8 12 12 12 8

Back Propogation

True

Learning rate

0.5

Momentum

0.1

Growth rate

0.5

run length:35

	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31	32	33
avg	0.1054728	0.1185876	0.1258131	0.130587	0.1506598	0.2122319	0.4581497	0.8586704	1.20199	1.170219	0.9763492	1.176481	0.938668	0.9572123	0.902207	1.133812	1.098755	0.9892185	1.5469	1.384855	1.156185	0.8816531	1.155313	1.370298	0.9086894	1.091801	0.9211531	1.064359	1.445552	0.9272204	1.219871	1.281623	1.345069	1.290509
max	0.1787508	0.3422044	0.4880244	0.4869769	1.047815	1.288545	1.504831	1.27506	1.43819	1.796796	1.367374	1.229414	1.067725	1.265468	1.230431	1.216018	1.179089	1.173458	1.681361	1.496624	1.225676	1.205523	1.230803	1.436737	1.25253	1.174299	0.9801109	1.207027	1.711045	1.16089	1.435449	1.345895	1.40991	1.366372

Population Size

100

Run Length

20

Do Training

False

Do Hybrid Training

True

Do Competitive Run

False

Do Adaptive Training

True

Inputs

hasTarget,targetLeft,targetRight,targetNorth,targetSouth,dirToTarget,distToTarget,wall,sensor0,sensor1,sensor2

Outputs

left,right,up,down,run,sensordir0,sensordir1,sensordir2

Hidden Layers

8 12 12 12 8

Back Propogation

True

Learning rate

0.5

Momentum

0.1

Growth rate

0.5

and a run w/another hidden layer

run length:33

	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31
avg	0.1104814	0.136816	0.1908503	0.3284849	0.6684353	1.282734	1.302261	1.206301	1.358841	1.126256	1.397736	1.246604	1.323351	0.8909027	1.122478	1.108233	1.631226	1.291764	1.220225	1.21547	1.299522	1.295794	1.360445	1.006665	1.005265	1.285506	1.723688	0.9848436	0.9455433	0.9313265	1.148989	1.272345
max	0.2871341	1.214576	1.067269	1.103958	1.354696	1.475417	1.366202	1.234541	1.390623	1.173298	1.426518	1.764977	1.396152	0.9108025	1.144939	1.133509	1.90503	1.320514	1.246103	1.239381	1.498761	1.339491	1.46306	1.135313	1.030648	1.328566	1.772488	1.166132	0.9746216	1.090431	1.179814	1.307164

Population Size

100

Run Length

20

Do Training

False

Do Hybrid Training

True

Do Competitive Run

False

Inputs

hasTarget,targetLeft,targetRight,targetNorth,targetSouth,dirToTarget,distToTarget,wall,sensor0,sensor1,sensor2

Outputs

left,right,up,down,run,sensordir0,sensordir1,sensordir2

Hidden Layers

8 12 12 12 8

Back Propogation

True

Learning rate

0.5

Momentum

0.1

Growth rate

0.5

Revisiting this whole neural net thing. Did another, longer run after tweaking a bunch of things (mostly removing cruft from the side-view test, like jumping and "grounded" and such).

run length:28

	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26
avg	0.1111835	0.1175151	0.1161713	0.1256572	0.1368807	0.156331	0.1656164	0.2083951	0.2679169	0.3795652	0.4956123	0.6501181	0.561227	0.763023	0.626345	0.684875	0.9196341	0.7514365	0.8885059	1.065271	0.9615093	0.809881	0.744675	1.001024	0.838229	0.8915754	0.8473446
max	0.2450459	0.2755109	0.2239121	0.4235514	0.4872672	0.4836805	0.9154767	1.048801	0.8001958	0.8589709	1.126522	1.106779	0.9452468	1.010383	0.7943439	1.048769	1.18526	0.9457676	1.149609	1.234839	1.202767	0.9807028	1.026631	1.079651	0.8743093	0.9314666	0.9513353

Population Size

100

Run Length

20

Do Training

False

Do Hybrid Training

True

Do Competitive Run

False

Inputs

hasTarget,targetLeft,targetRight,targetNorth,targetSouth,dirToTarget,distToTarget,wall,sensor0,sensor1,sensor2

Outputs

left,right,up,down,run,sensordir0,sensordir1,sensordir2

Hidden Layers

8 12 12 8

Back Propogation

True

Learning rate

0.5

Momentum

0.1

Growth rate

0.5