d7422cd99bMoved training files to separate folder in preparation of making training_files a submodulejudsonupchurch2025-05-13 01:40:59 +0000
1f00ca4da4Inverted pendulum report added. Should be finishedjudsonupchurch2025-04-06 22:23:15 +0000
89c06c5c42Fixed a huge issue where base loss centroid convergence was not using 'one' as its reference loss function, resulting in opposite trends.judsonupchurch2025-04-06 20:52:50 +0000
8155b0f7aeBasic report added. Report missing significant number of graphs and conclusionsjudsonupchurch2025-04-01 18:57:36 +0000
a1c931480fChanged linear regression to not account for log axesjudsonupchurch2025-04-01 00:12:36 +0000
867553353bGenerated convergence plots for 'centroid' parameters such as t_median, t_mean, R (later/early).judsonupchurch2025-03-31 21:50:30 +0000
4d90689a60Generated plots for time_weighting, time_weighting_learning_rate_sweep, and base_loss_learning_rate_sweepjudsonupchurch2025-03-30 22:22:10 +0000
bfa0f3fb02Trained additional base loss functions for lr sweep. Generated data for time weighting. Generated composite epoch evolution plotsjudsonupchurch2025-03-29 19:23:52 +0000
aa34bfac8cBase loss learning rate sweep training done. Made composite plots for epoch evolution of different time weightingsjudsonupchurch2025-03-26 04:20:07 +0000
a401ca3f59Done re-training max_normalized time weighting (with a minimum weight of 0.01) and time_weight_learning_sweep). Started work on base loss function trainingjudsonupchurch2025-03-12 22:33:37 +0000
eb71ab0de9Finish training with mirrored weights. Plot max normalized with mirror weightsjudsonupchurch2025-02-25 23:27:49 +0000
c89998da28Fix issues where the initialized controllers could be different. Created a controller_base.pth that is used for all controller initializationjudsonupchurch2025-02-24 02:28:16 +0000
8be7ad97a8Plotting theta across epochs for the different loss functionsjudsonupchurch2025-02-22 23:45:53 +0000
b9212e5a52Average normalized results across epochs have been plottedjudsonupchurch2025-02-19 03:06:48 +0000
8f92ce3ee1Restructure files. Changed weight functions to be normalized from 0-1ns to always be normalized from 0-1 (aka max normalization). Also updated average normalizationjudsonupchurch2025-02-18 18:45:15 +0000
28c5d14fe8Plotted controller max normalzied across epoch. Also training average normalizedjudsonupchurch2025-02-18 00:40:29 +0000
071669696bCompleted training for max normalized comparisonjudsonupchurch2025-02-17 18:47:05 +0000
3614b66aeeNormalization of loss functions based on max weightjudsonupchurch2025-02-17 02:54:56 +0000
3fc78d4508Looking at the different controllers as they trainedjudsonupchurch2025-02-16 00:03:29 +0000
judsonupchurch