Showing posts with label reinforcement. Show all posts
Showing posts with label reinforcement. Show all posts

2/27/09

#97: Types of Training




I have been a trainer for four years now. In that time, I have learned a lot about what it means to train and how you can go about doing it. I have also had the awesome opportunity to use a variety of training methods with our marine mammals. This blog post describes the different training methods that I have used and gives video examples of the behaviors that have come from them. It is pretty amazing when you open your eyes to the limitless ways you can train a behavior. Plus it's a lot of fun!



The training method that I use most often is called Shaping. By shaping a behavior, you are basically breaking it down into steps called approximations and you reinforce each step that the animal reaches. This allows you to build up a behavior and "shape" it as you go. Think of climbing a flight of stairs ... If you want to get to the eighth step, you have to walk up steps one through seven first. Shaping can be broken down even further into categories. The two shaping categories that I have used are successive approximations and selective reinforcement.



This above video is an example of Successive Approximations. Reggae is demonstrating his sink spin behavior. Since there are several components to the behavior, Reggae has to learn to spin in a circle, then to sink before putting the two behaviors together. In the video, you will see how he was trained to sink to successively deeper levels in the water column. Once that part was completed, he was given the spin signal which he was already trained to do at the surface. This requires a bit of abstract learning on his part since he was never asked to spin underwater but he can handle it! I can't take credit for training this one but it is a perfect example of the approximations needed to build a behavior.

Another category that is part of the shaping method of training is Selective Reinforcement. Before they headed off for their vacation in New York, I worked with our northern fur seals Cordova and Ursula. Both of them were quick studies and selective reinforcement was a new and exciting way for me to train behaviors with them. To train using this method, you basically wait until the seal offers something that you want. Then, you reinforce only that movement.

An example of this is Cordova's spin behavior. To start the training of this behavior, I gave Cordova (pictured at left) the spin signal (which she had never seen before) and then waited until she turned her head ever so slightly. From there, I continued to reinforce her head movement as long as it was past the point she reached the last time. As her head would turn, her body would follow and she would begin to turn in a circle.

By using this technique, I was able to train Cordova to spin in two days! It also gave me an opportunity to fine tune my bridging skills. It is really important to give a clear message to the seal so the accuracy of your bridge is crucial. This is a also a great training method to use when trying to introduce an animal to new surroundings. So stay tuned for the return of our northern fur seals because I am sure we will be using selective reinforcement as a very helpful training tool when introducing them to their brand new exhibit!

Shaping is also a really helpful way to rework an old behavior that has broken down over time. There are a number of reasons a behavior breaks down or no longer meets criteria. The seal may have been reinforced for a lesser version of what was originally required. Sometimes other trainers give a different version of the signal that the seal may not recognize. Reggae's dance behavior became progressively lower and slower. Below is a video of of the behavior.



Right now, he doesn't extend his flippers very far out of the water and he also moves verrrrrry slooooowly. Through shaping, I am hoping to sharpen this behavior so that he dances the salsa (like Cayenne in the video) rather than a waltz. Stay tuned for an update ...

Another training category that has been really fun to use with Reggae is Capturing. This is a training technique where you capture a behavior that the animal offers on their own. Reggae exhibits many different fun and energetic behaviors during the breeding season. Reggae would often offer these different behaviors after training sessions so I would wait on exhibit with a few fish and feed him when he did the desired behavior. In the video below, you can see him offering all sorts of behaviors in an attempt to get reinforced. The more he was reinforced for doing a specific behavior, the more he offered it. Then I picked a word or a signal to go with the behavior and transferred it into training sessions. This video shows a couple sessions where I am trying to capture Reggae's underwater bubbles. You can see him starting to understand what I am looking for because he offers behaviors other than underwater bubbles less and less as we go along.




The behaviors Reggae offers during breeding season usually involve blowing bubbles, slapping the water and quick body movements. Over the past two years I have been able to capture three of these behaviors and pair then with a signal. This video shows all the cool behaviors that Reggae has come up with. Check it out! :)

Some of the many captured behaviors from Reggae the Atlantic harbor seal at the New England Aquarium.



So the sky is the limit with the types of training we utilize with our animals but it always comes down to one main theme ... positive reinforcement. Make it positive for the animal and for the trainer and you are bound to have success!


Thanks to all our volunteers who helped me videotape Reggae's behaviors. I don't know what we would do without you!


~Patty

subscribe




11/20/08

#66: What does that mean?



Alright, we get it. Sometimes we use terms that make sense to us, but not necessarily to you. So here are a few of the terms that you might find useful when we talk about training:


Signal: Any touch, word, sound or visible gesture that cues the animals to do a particular behavior. For example: saying the word "Wave" directs a seal to move their flipper back and forth. The word wave is a verbal signal.

Putting our hands out and moving from side to side is a gesture that directs the seals to dance. We call this signal a hand signal.

Touching the top of a seal's nose is a signal that cues them to exhale. The touch is called a tactile signal.

Here you can see Jenny giving Lana the hand signal to dance - notice that Lana is raising her left flipper out of the water to mirror Jenny's right hand. Check out this video of dancing behavior.


Bridge: A signal that is used to indicate the precise instant that the seal did something correctly. When the seal hears or feels the bridge, they immediately stop what they're doing and look to the trainer for fish or other reinforcement. It "bridges" the time between the behavior and reinforcement. Like a signal, a bridge can be visual, auditory or tactile.

Some of the bridges that we use include saying the word "Good," a short blow on a whistle, or two gentle taps on the seal's body.

Here is Baranov using his bridge...

...and Paul is using his whistle bridge.





Reinforcement: Anything the animal wants, needs or wants more of. Here at the Aquarium, our main reinforcement for the seals is fish, but it can also be a favorite toy, a scratch on the chin or any number of things. It's good to have multiple types of reinforcement--just like people, each seal has it's own likes and dislikes which can change over time. Also, not all seals like the same thing. One seal might enjoy a scratch under the chin while another one may not want to be touched.


Here, Jenny is using fish to reinforce Lana...



...but sometimes Lana likes to have her neck scratched.




Target: Any object the animal is taught to touch with their nose, flipper or other part of their body. Targets

provide a focal point and can be used to train any behavior. The seals can be comfortable targeting for several minutes during ultra sound or blood sampling. Gradually raising the height of a target over the water helps teach a seal where to jump. Here at the Aquarium, we use three types of targets: a hand target, a target pole with a bead or ball target on the end, and a name target.

Each seal has a specific object for their name target. At the start of each training session, the seals will look or listen for their particular target. When they find it, they know where in the exhibit their session will take place and which trainer will do their training session.

Here you see Smoke is targeting on a trainers hand while Reggae targets on his name target (a plastic tire) and Amelia targets on her name target (a black plastic disc with slits in it). If you look in the background you will also see examples of other name targets--notice how they are all different.

In this video you can see how they all work together:


When I want to ask Chuck and Cayenne to go to their Name targets, I point to the targets (hand signal) while saying the word "Target" (verbal signal). Once Chuck and Cayenne touch their nose to their individual targets, I will say the word "Good" (bridge) and feed them some fish (reinforcement).

Using these four tools correctly is what has allowed us to train so many behaviors. It's fun. Try it with your friends!

subscribe