Fading rewards using clicker

Discussion in 'Clicker Training' started by Widgeon, Nov 3, 2014.

  1. Widgeon

    Widgeon Registered Users

    Joined:
    Nov 4, 2013
    Messages:
    81
    I found an archive article, posted by Pippa, about clicking and NOT treating.
    I read the follow up posts but am not totally clear on the end thoughts.

    I use a range of hi value treats from sausage and bacon flavour treats and a low value, part of his diner portion kibble. I have been fading the hi's and increasing the lows and starting to miss food treats altogether occasionally, it seems most people will ALWAYS treat after a click.
    How do you 'pro-clickers' fade treats?
    I appreciate a treat doesn't have to be food but he's not greatly fussed by praise and headrubs (I don't think??)
    Do I swap a low with a heart felt "good boy" or just move the difficulty up?
     
  2. snowbunny

    snowbunny Administrator Forum Supporter

    Joined:
    Aug 27, 2014
    Messages:
    14,763
    Location:
    Andorra and Spain
  3. Widgeon

    Widgeon Registered Users

    Joined:
    Nov 4, 2013
    Messages:
    81
    Re: Fading rewards using clicker

    That was the very thread that I was reading!

    I didn't feel completely confident of the end reply, and wondered what people do in practice.
     
  4. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    22,884
    Re: Fading rewards using clicker

    There are some discussions about not always following a click with a treat. There is an article here:

    http://www.patriciamcconnell.com/theotherendoftheleash/click-and-always-treat-or-not

    I always follow a click with a treat. I am still thinking about it, but for now I always treat. I'd say if you are starting off, and your dog is new to clicker training, always treat after a click.

    So in this way, I never fade rewards when I am clicker training. I clicker train something, then I put it on cue. Then it is finished.

    This is, I think, a different set of considerations from whether you continue to "pay" with treats to maintain trained behaviour.
     
  5. Oberon

    Oberon Moderator Forum Supporter

    Joined:
    Mar 6, 2013
    Messages:
    15,552
    Location:
    Canberra, Australia
    Re: Fading rewards using clicker

    I always treat after a click and I never completely stop treating for any behaviour. However I do increase the standard required for a click and treat, over time (eg behaviour has to be longer, more precise, faster, or becomes part of a chain of behaviours). I will regularly pay up though - removing all rewards will lead to the behaviour disappearing (unless it is an inherently rewarding behaviour that your dog has come to love for its own sake like the opportunity to retrieve).

    When teaching a young dog you should be looking for every opportunity to click and treat (gradually shaping the behaviour towards your goal, eg a straight sit or a fast down or a fast recall in a distracting situation). The biggest problem I see with young dogs is handlers failing to notice good attempts from their dog and therefore failing to reward and therefore reducing the chances the the dog will try again in future.

    Praise can become valuable to a dog but usually only because it's become associated with something more important like food. Praise is not a 'primary reinforcer', meaning that it's not fundamentally important to the dog. Praise is important, but you still need to regularly back it up with food :)
     
  6. snowbunny

    snowbunny Administrator Forum Supporter

    Joined:
    Aug 27, 2014
    Messages:
    14,763
    Location:
    Andorra and Spain
    Re: Fading rewards using clicker

    Rachael, some wise words there. At the risk of going a bit off topic, could you give me a brief idea of how you would work on improving speed? For example, Willow is getting very good at hand targeting and will do it from a good distance now, but she always ambles over rather than putting any effort into it. Any tips on making her a little more lively? Thanks!
     
  7. Oberon

    Oberon Moderator Forum Supporter

    Joined:
    Mar 6, 2013
    Messages:
    15,552
    Location:
    Canberra, Australia
    Re: Fading rewards using clicker

    One thing you could try is having someone hold her, then you really rev her up in a very excited way. When she really wants to get to you say your target cue and have the other person release her (she should be straining to get to you). Big reward. Be excited!!!!! Don't make the reward contingent on anything other than speed to get to you to touch your hand (so don't require a sit or anything additional). Also, make sure she is hungry and wanting attention and so do it before dinner :) You can even make dinner the reward if you've got it ready to slam down on the floor (although some dogs get so focused on the presence of their dinner that they can't concentrate on anything else).

    Whatever speed she offers you, make that your baseline and don't reward anything less. I'd only do two of these in a day, maximum.
     
  8. snowbunny

    snowbunny Administrator Forum Supporter

    Joined:
    Aug 27, 2014
    Messages:
    14,763
    Location:
    Andorra and Spain
    Re: Fading rewards using clicker

    [quote author=Oberon link=topic=8591.msg121326#msg121326 date=1415033733]
    Also, make sure she is hungry
    [/quote]

    This is a labrador we're talking about ;)

    Thanks for the advice. I'll try it tomorrow.
     
  9. editor

    editor Administrator

    Joined:
    May 10, 2011
    Messages:
    6,412
    Re: Fading rewards using clicker

    I think there may be some differences over the way we are using the word 'fade'. By fade rewards I mean ask for more behaviours, or longer (duration) behaviours before offering reinforcement. Or offering lower value rewards for some behaviours. Or a combination of the two. Essentially, the dog is being paid less, for doing more. In practice, in gundog training the dog is actually getting higher value (to the dog) rewards as training advances, so 'fading' tends to refer to the use of food, rather than reinforcements in general.
     
  10. BonkersAsConkers

    BonkersAsConkers Registered Users

    Joined:
    Dec 23, 2015
    Messages:
    7
    I know some people don't always treat after a click but I always do.
    To me the whole power of the click comes from the treat after.
    You might get away with missing one here and there but there will come a point where click starts to mean "You MIGHT get a treat" and I don't want that to happen.
    I'd rather teach a "Keep going" signal so for instance - Cue-dog begins behaviour eg heel work-"good" (=keep doing that, you're well on the way to getting a treat)-click/yes-treat.
    Then I fade the clicker once behaviour is reliable and on cue and start to vary the reinforcement rate.
    Also "Click is always followed by a treat" is easier to remember for beginner clickerererers than "Usually a click is followed by a treat but not always".
     
  11. SteffiS

    SteffiS Supporting Member Forum Supporter

    Joined:
    Aug 14, 2015
    Messages:
    887
    Location:
    Fife, Scotland
    I've been working hard on Ripple walking without pulling and I click at regular intervals all the time he's walking nicely, however I only treat at irregular intervals. I'm wondering whether this is wrong as whenever I give a treat with the click he starts to pull again as if he thinks the treat is a signal the good behaviour is no longer required.
     
  12. Oberon

    Oberon Moderator Forum Supporter

    Joined:
    Mar 6, 2013
    Messages:
    15,552
    Location:
    Canberra, Australia
    I'd try treating with every click (I do the same as BonkersAsConkers and always reward after a click) but:
    (a) gradually extend the duration between clicks. Basically, expect a longer duration of nice walking before your dog earns a click. 'Raise the price' of a click.
    (b) still throw in there some clicks for much shorter periods of good walking - don't make it harder and harder and harder
    (c) continue to use the 'stop' or 'u-turn' if pulling or poor walking happens. Remove the reward (going forward) for poor walking. Be 100% consistent about this
    (d) practice the above in new locations, going back to basics (as if teaching for the first time) if you need to combat the distraction of the new environment.
    (e) for familiar environments where you know your dog can do a pretty good job, switch to less interesting rewards (like kibble). Save the really good treats for more difficult situations or really excellent performances.
     
  13. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    22,884
    Charlie used to do this. Now I click, stop, give him his treat, expect him to look at me after the treat and the reward for that is to continue the walk. The stopping, eating, looking breaks the "got my treat, now I'll pull again" thing. I think it's a natural thing for him to do, as in lots of other situations the click is the release - as in "you've done what I wanted".
     
  14. SteffiS

    SteffiS Supporting Member Forum Supporter

    Joined:
    Aug 14, 2015
    Messages:
    887
    Location:
    Fife, Scotland
    JulieT I can see now how Ripple could mistake the treat as a signal for the end of the behaviour, I just thought he was being awkward :). I shall try to incorporate your method to correct this.

    Oberon I know I need to work at the click and treat, I've bought the book recommended on here but not had time to study it with all the Christmas stuff to get sorted. Plus I'm also having trouble with Ripple's adolescent behaviour as he's started biting again so I seem to be battling lots of things at the moment.
     

Share This Page