Update README.md

README.md

@@ -25,12 +25,21 @@ model-index:
   This is a trained model of a **Reinforce** agent playing **CartPole-v1**.
   To learn to use this model and train yours check Unit 4 of the Deep Reinforcement Learning Course: https://huggingface.co/deep-rl-course/unit4/introduction
 
+
+  # ***Project Information***
+
+
   **Policy-based learning** is directly approximating π without having to learn a value function- Our objective then is to maximize the performance of the parameterized policy using gradient ascent.
   TL;DR: Having the cart learn to balance the pole via optimizing π for the best output; *the pole not falling over*. 
   This method of learning skips over using a value function like Q-learning does, allowing an immediate improvement in the next iteration instead of having to calculate and approximate tables and numbers for a new action, as Q-learning does.
+  
 
   This specific CartPole model only has 500 training timesteps- the average is 1000, which is the reason why the cart struggles so much with balancing the pole in the video; it has not trained enough for it.
   A model trained with 1000 timesteps is successful in balancing the pole, and the more training steps a model has, the more accurate its result is, like when you play a really hard level in a video game over and over, it eventually gets easier.
-  However, the more timesteps a model has, the longer it takes to train and render- 1000 timesteps takes 10-15 minutes to load, and the time only increases the more training timesteps are inputted. 
+  However, the more timesteps a model has, the longer it takes to train and render- 1000 timesteps take 10-15 minutes to load, and the time only increases the more training timesteps are inputted. 
+
+  Here -https...- is a video of it working with 1000 timesteps, and here -https...- is one with 2000 *(links will be inserted soon)*
+
+