diff --git a/newton-4.00/sdk/dBrain/ndBrainAgentContinueVPG_Trainer.h b/newton-4.00/sdk/dBrain/ndBrainAgentContinueVPG_Trainer.h
index 1c4375377a..e571a15711 100644
--- a/newton-4.00/sdk/dBrain/ndBrainAgentContinueVPG_Trainer.h
+++ b/newton-4.00/sdk/dBrain/ndBrainAgentContinueVPG_Trainer.h
@@ -270,7 +270,7 @@ ndBrainAgentContinueVPG_Trainer<statesDim, actionDim>::ndBrainAgentContinueVPG_T
 
 	m_baseLineValueOptimizer = new ndBrainOptimizerAdam();
 	//m_baseLineValueOptimizer->SetRegularizer(hyperParameters.m_regularizer);
-	m_baseLineValueOptimizer->SetRegularizer(ndBrainFloat(1.0e-5f));
+	m_baseLineValueOptimizer->SetRegularizer(ndBrainFloat(1.0e-4f));
 
 	m_baseValueWorkingBufferSize = m_baseLineValue.CalculateWorkingtBufferSize();
 	m_workingBuffer.SetCount(m_baseValueWorkingBufferSize * hyperParameters.m_threadsCount);