Progression to a random reward scheme... - German Shepherd Dog Forums
Reply
 
LinkBack Thread Tools
post #1 of 52 (permalink) Old 11-10-2013, 06:09 AM Thread Starter
Member
 
Join Date: Feb 2013
Location: Sydney Australia
Posts: 77
Progression to a random reward scheme...

I was wondering whether anyone could share their own methods of reward progression in training (particular actions as well?), preferably from those who use marker or clicker training. Because it's all well and easy to teach a dog to sit, then sit for duration, sit under distraction for duration etc. But I have the most difficulty with reward progression. My current reward progression is based on Michael Ellis' scheme:

constant > variable > random

But I still am struggling to understand what I should be doing after I deliver the reward cue mark, in a variable-off or random-off segment. It just seems awkward - for instance, if I'm rewarding every second turn, what do I do when I deliver my reward cue mark on a non-reward delivering round? My resources don't teach this is any great detail - so I praise my dog, but naturally she's frantically searching for her reward (food).I'm worried I may be counter-conditioning the mark and causing extinction. She has an absolutely manic and I mean MANIC food-drive.

So how is a non-reward turn training segment structured in your training sessions? Do any of you have different schemes for reward progression? Do you guys agree with the scheme I have described? Your thoughts would be greatly appreciated. Particularly, those of senior members and IPO/Schutzhund folks.
Alwaysaworkingdog is offline  
Sponsored Links
Advertisement
 
post #2 of 52 (permalink) Old 11-10-2013, 12:53 PM
No Stinkin' Leashes Moderator
 
Cassidy's Mom's Avatar
 
Join Date: Mar 2003
Location: SF Bay Area
Posts: 31,940
Quote:
Originally Posted by Alwaysaworkingdog View Post
But I still am struggling to understand what I should be doing after I deliver the reward cue mark, in a variable-off or random-off segment. It just seems awkward - for instance, if I'm rewarding every second turn, what do I do when I deliver my reward cue mark on a non-reward delivering round?
The marker ALWAYS means that a reward is coming, so if you're rewarding randomly, you're only going to mark the repetitions that you intend to reward. You can continue to praise, you just don't use your marker.

-Debbie-
Cava 1/6/18
Keefer 8/25/05-4/24/19 ~ The sweetest boy
Halo 11/9/08-6/17/18 ~ You left pawprints on our hearts
Dena 9/12/04-10/4/08 ~ Forever would have been too short
Cassidy 6/8/00-10/4/04
Cassidy's Mom is offline  
post #3 of 52 (permalink) Old 11-10-2013, 01:08 PM
Elite Member
 
Merciel's Avatar
 
Join Date: Apr 2013
Location: Philadelphia
Posts: 1,574
For me, "random" never really means "random," it means the best out of a progressively higher number of performances. In the learning/shaping phase, every "more correct than the last" response might get rewarded. In the polishing phase, it's best out of 3, or best out of 5, or best version of whatever specific issue I'm trying to perfect.

For example, right now I'm working on tightening Pongu's Fronts during formal retrieves, so today's criteria is that every Front within 4 inches of my toes gets rewarded, and every Front that's farther away gets a little bit of praise and then we reset the exercise to do it better next time.

I do not click (or mark) performances that aren't going to be rewarded. The ones that are good, but not good enough, get a little bit of praise and then we go back and do it again.

ARCHMX TDCH Pongu the Insane, CD-C, RE, RL1X6, RL2X5, RL3X2 (GSD mix, b. Apr 2010)
Crookytail the Tigerwuff, RL1, ITD (Akita mix, b. Jan 2011)
Merciel is offline  
Sponsored Links
Advertisement
 
post #4 of 52 (permalink) Old 11-12-2013, 01:35 AM Thread Starter
Member
 
Join Date: Feb 2013
Location: Sydney Australia
Posts: 77
Quote:
Originally Posted by Cassidy's Mom View Post
The marker ALWAYS means that a reward is coming, so if you're rewarding randomly, you're only going to mark the repetitions that you intend to reward. You can continue to praise, you just don't use your marker.
I'm still not sure how I do that though. You wouldn't be able to describe how you would conduct a training session on a random reward scheme would you? My main problem is that my dog will not break the action until I give the reward mark. Now I don't want to say something else, because that will probably just confuse the dog.

But if I'm, say, teaching my dog to sit. She sits for duration, under distraction and has gone through a correction phase AND THEN I start to change my reward scheme, first to, let's say, a variable scheme. So I decide that I'm going to reward her variably, every 2nd sit. So I give her a commencement mark and we begin;

I tell her "sit"

she sits

then what? I kind of feel jarred, because I'm so used to using 'yes' as a "the action is over, good job, come get your reward" kind of mark. How do I get her to move out of sit and on to the next repetition? I can't tell her "sit", because she's obviously already sitting. I can't say to her "yes", because I'm only rewarding every second repetition. Should I just use leash pressure and pull her out of the sit? I'm quite confused....
Alwaysaworkingdog is offline  
post #5 of 52 (permalink) Old 11-12-2013, 01:58 AM
Banned
 
Join Date: Jun 2013
Posts: 1,158
after you put her in a sit give her another command. do you want her to stand? do you want her to heel by your side and walk? do you want her to lay down? do you want her to know that training is over? if your dog is sitting under distraction and distance then just give her a here command and have her run to you. you can add a few commands together before rewarding. i mark with yes but my command to release my dog from training is "free". my dog either has to wait for another command or the word free before moving.

another thing you can do is teach her the command "good". when she sits you can praise her and sometimes pet her and say good or whatever word you choose. she'll learn that the word good means shes doing it right but it doesnt release her from the command. this is how i add distance and duration to all my commands. i say good every few seconds so my dog knows its doing the command correctly and then at the end i will use my marker word and the dog is released and gets the reward. good just means the reward is eventually coming.
boomer11 is offline  
post #6 of 52 (permalink) Old 11-12-2013, 01:59 AM Thread Starter
Member
 
Join Date: Feb 2013
Location: Sydney Australia
Posts: 77
Quote:
Originally Posted by Merciel View Post
For me, "random" never really means "random," it means the best out of a progressively higher number of performances. In the learning/shaping phase, every "more correct than the last" response might get rewarded. In the polishing phase, it's best out of 3, or best out of 5, or best version of whatever specific issue I'm trying to perfect.

For example, right now I'm working on tightening Pongu's Fronts during formal retrieves, so today's criteria is that every Front within 4 inches of my toes gets rewarded, and every Front that's farther away gets a little bit of praise and then we reset the exercise to do it better next time.

I do not click (or mark) performances that aren't going to be rewarded. The ones that are good, but not good enough, get a little bit of praise and then we go back and do it again.
But hypothetically, say that Pongu's Fronts were perfect, how would your random reward scheme look? You mentioned you're praising, so that's kind of like using your "duration/good" mark as a kind of reward or are you just giving general praise, does praise become a reward? but just an informal one?

In terms of exercises with very clearly defined beginnings and ends, such as sits and downs, how does your random scheme play out? If you have/had one. Is it the same as in retrieve? Does your praise come to signal the end of the action, in the eyes of your dog? Because then I won't use my duration mark as a praise, because that'll counter-condition my mark.

Furthermore, what happens BEYOND the random scheme, waaaaayyyy down the road. Once you guys have your commands down (and I'm talking about your bread and butter ones that your dog has mastered), how do you keep the dog motivated and avoid extinguishing the behaviour? Is it just about rewarding randomly? How does this play out in everyday life? Say you're down at the park and your dogs prong falls off or something and you put him/her in a down, or something happens and you have to use a command - how often are you rewarding that behaviour? how often do you do sessions for that behaviour? is it a matter of routinely having sessions for upkeep of a certain command or is that not necessary as long as the dog is keeping up quality behaviour?

Even the best training resources seem to really neglect the tail end of marker training - you don't need to answer all the questions, because I just read back and realised how many there are haha

just if you or anyone has the time I could really use some help.

Regards
Alwaysaworkingdog is offline  
post #7 of 52 (permalink) Old 11-12-2013, 02:10 AM Thread Starter
Member
 
Join Date: Feb 2013
Location: Sydney Australia
Posts: 77
Quote:
Originally Posted by boomer11 View Post
after you put her in a sit give her another command. do you want her to stand? do you want her to heel by your side and walk? do you want her to lay down? do you want her to know that training is over? if your dog is sitting under distraction and distance then just give her a here command and have her run to you. you can add a few commands together before rewarding. i mark with yes but my command to release my dog from training is "free". my dog either has to wait for another command or the word free before moving.

another thing you can do is teach her the command "good". when she sits you can praise her and sometimes pet her and say good or whatever word you choose. she'll learn that the word good means shes doing it right but it doesnt release her from the command. this is how i add distance and duration to all my commands. i say good every few seconds so my dog knows its doing the command correctly and then at the end i will use my marker word and the dog is released and gets the reward. good just means the reward is eventually coming.
Thanks for the reply boomer. Those are good ideas. Linking commands just seems to confuse me more though because I'd have to have all these actions trained simultaneously. I don't want her to start confusing down and sit by lumping them in the same training session, as that seemed to happen with her when I was initially training and I had to stop and rethink what I was doing. So you're saying that rather then reward her randomly, I should just link behaviours and reward the last one? Or not reward the last one? I don't want to use my end mark "ok" because I don't want the session to end, I want to keep going and teach her that she's now only being rewarded every second go.

I use 'good' as a duration mark already, but she's gotten to the stage where she doesn't need it.
Alwaysaworkingdog is offline  
post #8 of 52 (permalink) Old 11-12-2013, 02:24 AM
No Stinkin' Leashes Moderator
 
Cassidy's Mom's Avatar
 
Join Date: Mar 2003
Location: SF Bay Area
Posts: 31,940
Quote:
Originally Posted by Alwaysaworkingdog View Post
I don't want to use my end mark "ok" because I don't want the session to end, I want to keep going and teach her that she's now only being rewarded every second go.
No time for more right now, I just wanted to quickly respond to this. If you're using "ok" to end your training sessions, what are you using as your release word? It doesn't matter what words you use, I just want to know if you have a word to release her from whatever command you've given.

For me, "ok" releases the dog from the command, so if the dog is in a sit/stay or a down/stay, they can get up. If I'm ending a training session I use a different cue, such as "all done".

Also, one very good way to transition away from food rewards for easy behaviors that are learned and generalized (such as sit), is to use the command in day to day life to get the dog what they want. For example, you can have her sit until released while you put her food bowl down, and getting to eat is the reward, or sitting and looking at you to "make" you open the door and let her out or in, before putting on her leash for a walk, or before throwing a ball or letting her take a tug toy. She's still being reinforced for the sit, but you're not having to mark and reward with a treat each time.

-Debbie-
Cava 1/6/18
Keefer 8/25/05-4/24/19 ~ The sweetest boy
Halo 11/9/08-6/17/18 ~ You left pawprints on our hearts
Dena 9/12/04-10/4/08 ~ Forever would have been too short
Cassidy 6/8/00-10/4/04
Cassidy's Mom is offline  
post #9 of 52 (permalink) Old 11-12-2013, 03:07 AM Thread Starter
Member
 
Join Date: Feb 2013
Location: Sydney Australia
Posts: 77
Quote:
Originally Posted by Cassidy's Mom View Post
No time for more right now, I just wanted to quickly respond to this. If you're using "ok" to end your training sessions, what are you using as your release word? It doesn't matter what words you use, I just want to know if you have a word to release her from whatever command you've given.

For me, "ok" releases the dog from the command, so if the dog is in a sit/stay or a down/stay, they can get up. If I'm ending a training session I use a different cue, such as "all done".

Also, one very good way to transition away from food rewards for easy behaviors that are learned and generalized (such as sit), is to use the command in day to day life to get the dog what they want. For example, you can have her sit until released while you put her food bowl down, and getting to eat is the reward, or sitting and looking at you to "make" you open the door and let her out or in, before putting on her leash for a walk, or before throwing a ball or letting her take a tug toy. She's still being reinforced for the sit, but you're not having to mark and reward with a treat each time.
Hence why I started this thread haha

maybe that's my problem. My marks are:

start session: ready
reward: yes
correction: no
duration: good
end session: ok

So I don't have a mark similar to your "ok". I suppose it would be "yes", but that indicates to her that she's getting a reward. Because formerly, when I was working by rewarding each repetition, "yes" served that purpose.

Are you saying I need to create another mark for when I progress to a variable scheme? So that on the non-rewarding rounds, I can use that?

Like say I make that mark "done" and I'm moving to a every 2nd repetition variable scheme:

# sit > done

# sit > yes (reward)

# sit > done

# sit > yes (reward)

^ is that kind of how I should structure it?

How do I charge that mark? Do I combine it with leash pressure or something? Because if I introduce a new mark, out of the blue, she won't know what I'm saying and she won't "release" herself....
Alwaysaworkingdog is offline  
post #10 of 52 (permalink) Old 11-12-2013, 03:30 AM
Moderator
 
David Winners's Avatar
 
Join Date: Apr 2012
Location: NW Ohio
Posts: 2,558
Sit-good
Down-good
Stand-good
Sit-yes

Shuffle and repeat.

David Winners

When a dog saves the life of a man, it becomes clear that partnership knows no bounds.

Fama - T.E.D.D. OEF XI-XII (GSD)(RIP)
Marshall - T.E.D.D. OEF XII-XIII (Lab)(SF EDD)
Lucian - Med Alert / Mobility SD (Cane Corso)
Pud - the old man (Pit x Lab)
Hank - DW Dog (Cane Corso)
David Winners is offline  
Sponsored Links
Advertisement
 
Reply

Quick Reply
Message:
Options

Register Now



In order to be able to post messages on the German Shepherd Dog Forums forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.

User Name:
Password
Please enter a password for your user account. Note that passwords are case-sensitive.

Password:


Confirm Password:
Email Address
Please enter a valid email address for yourself.

Email Address:
OR

Log-in










Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page



Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On

 
For the best viewing experience please update your browser to Google Chrome