wayne
Junior Hunter/worker
Posts: 34
|
Post by wayne on Feb 21, 2010 18:39:44 GMT -5
All, I don't know if anyone is interested in how I'm progressing with the so called "clicker" training of my dog. Its coming along fine. I made a couple of rough videos showing actual raw footage of her responding to my commmands. Not to bad for a 10 month old. Please keep in mind this is with no force applied. Totally volunteered behaviour, rewarded with treats. I don't usually have the goodies in my hand but forgot as I was trying to balance 2 dogs, camera, walking on a snow filled path etc and just plain forgot to put my hand back in my pocket. I didn't use the clicker as I unfortunately only have 2 hands, both of which were busy. But you fade the clicker out and use verbal or whistles or whatever after the dog knows the behaviour and is proofed with it. This first video I called her off of digging out a mouse. www.youtube.com/watch?v=iNkK6iOkwWkThis one is poor as I forgot to put the treat in my pocket and treated her too much I am phasing out the treats. www.youtube.com/watch?v=odAG0t51p5kThis one shows wait and here www.youtube.com/watch?v=nie4cgPiP44This one is another poor one, as I am confusing her. It shows me attempting to teach her OUT. I've since stopped this as I was not getting the behaviour that I wanted. www.youtube.com/watch?v=qcWgp8lE2nMMore of the same, I'm trying different things and am digging myself a hole. Back to basics. In this case, it's something called targeting www.youtube.com/watch?v=gx1CFPBy7t8I'm making all sorts of mistakes with the training. Katie doesn't seem to mind and transitions easily. Her Heeling,HERE, WAIT, SIT, DOWN are constantly improving. The next step is to start fading out the reward and substitute it with something else - praise? So far clicker training is very much similar to my idea of regular traditional training except that you get the dog to voluntarily do the action, then put a name to it after the dog gives you what you want consistently. Wayne
|
|
|
Post by brandon on Feb 22, 2010 18:19:21 GMT -5
why do you fade out the reward, do you not wont the dog to keep getting payed for her work is not better to very the reward but still pay here for working for you, what about a tug reward or a ball reward would that not be more of a reward than saying ( that a girl) could you not use her prey drive to reward the behaviour
|
|
wayne
Junior Hunter/worker
Posts: 34
|
Post by wayne on Feb 22, 2010 19:23:40 GMT -5
Hi Brandon, As you may be able to tell from the videos, I'm a rank amateur at this clicker training. I'm fading it out on a variable schedule (if you can call it that) and substituting praise words in place of the treats. She knows the commands, has proofed them in different venues and the treats are starting to lose their effectiveness. She is getting "paid" by having my approval which she is eager to get. Don't get me wrong, right now she is still getting treats but at a frequency of a lot less than I originally was giving to her. The schedule of treats is totally erratic (like the pay out at the casino). If I notice her attention starting to wander then I may treat more often with a combination of talking to her and food treats. Being a puppy she has a very short attention span, treats or not. Also note where I'm training her - mostly in the bush while walking on a trail. All sorts of distractions etc for her to investigate. What I am doing while walking home is I keep her on a shorter leash and she mostly walks at heel with very little surging ahead or lagging behind. This is not, I think, the recommmended method but it seems to work and she doesn't seem to mind.
Using a tug or ball is a lot slower and at this point I think counter productive as I'm also working with her at home in the house learning HOLD, OUT. I'm planning on using her prey drive for the retrieve which comes after the HOLD and OUT. She will be on a lead when I start insisting on the retrieve. Saying that she retrieves very nicely for me now. Not consistently but is willing depending on her (mood?,attitude?time of day? interest?)
Thanks for your interest and comments.
Wayne
Wayne
|
|
|
Post by brandon on Feb 23, 2010 8:32:05 GMT -5
Wayne I thought this might help with the clicker retrieve. I have used this before in the past and it is a easy way to teach your dog the idea before you go to force if you so choice that method, i know it is probably not what you are training for but a lot of the concepts are used in all types of dog training( Schutzhund-Training.com)
|
|
jax
Hunter/worker
Posts: 133
|
Post by jax on Feb 23, 2010 10:47:50 GMT -5
Hi Brandon, As you may be able to tell from the videos, I'm a rank amateur at this clicker training. I'm fading it out on a variable schedule (if you can call it that) and substituting praise words in place of the treats. She knows the commands, has proofed them in different venues and the treats are starting to lose their effectiveness. She is getting "paid" by having my approval which she is eager to get. Don't get me wrong, right now she is still getting treats but at a frequency of a lot less than I originally was giving to her. The schedule of treats is totally erratic (like the pay out at the casino). If I notice her attention starting to wander then I may treat more often with a combination of talking to her and food treats. Being a puppy she has a very short attention span, treats or not. Also note where I'm training her - mostly in the bush while walking on a trail. All sorts of distractions etc for her to investigate. What I am doing while walking home is I keep her on a shorter leash and she mostly walks at heel with very little surging ahead or lagging behind. This is not, I think, the recommmended method but it seems to work and she doesn't seem to mind. Using a tug or ball is a lot slower and at this point I think counter productive as I'm also working with her at home in the house learning HOLD, OUT. I'm planning on using her prey drive for the retrieve which comes after the HOLD and OUT. She will be on a lead when I start insisting on the retrieve. Saying that she retrieves very nicely for me now. Not consistently but is willing depending on her (mood?,attitude?time of day? interest?) Thanks for your interest and comments. Wayne Wayne Have you tried reward schedules?
|
|
wayne
Junior Hunter/worker
Posts: 34
|
Post by wayne on Feb 23, 2010 11:11:08 GMT -5
Brandon, Thanks for the link. It looks good. I'll watch the videos again when I need to so that I can remind myself that there is an end in sight. This is very similar to what I have planned to do, I may be able to use some of this. I'm trying to stay away from the force fetch and E collar. We are still working on the basics, but I am starting her on other portions of the eventual aim... to have her retrieve ducks, geese, upland game birds. She is extremely interested in birds. She is always the first to notice birds flying overhead. My GSD and myself were training with a Schutzhund club approx 6 yrs ago. I didn't like some of the methods that the club used, so I dropped out for a variety of reasons. My GSD is a well behaved, well trained dog as a result of this experience. Jax I don't know what a reward schedule is.
Thanks for both responses.
Wayne
|
|
jax
Hunter/worker
Posts: 133
|
Post by jax on Feb 23, 2010 11:47:27 GMT -5
Jax I don't know what a reward schedule is. Wayne I don't remember them all. My cousin Michael explained them too a crowd at a seminar I attended put on by a law enforcement group. All reward training is predicated on the strength and value of the reward offered to the dog for performance. Reward schedules enhance the reward and thus the performance. The two I do recall are: "extinction" which means that the dog is never rewarded for incorrect or no performance; and "continuous" which means that the dog is always rewarded for a correct response/performance. The schedule evolves to a single reward after two or more responses and include time feature rewards for performances like stays of a predetermined duration. If you like, PM me and I'll send you Michael's e mail address and he can explain it for you. I asked because there does not seem to be much emphasis on reward schedules in any of the clicker stuff I have read.
|
|
wayne
Junior Hunter/worker
Posts: 34
|
Post by wayne on Feb 23, 2010 13:46:01 GMT -5
Jax, I think that I know what you are talking about regarding reward schedule. 1:1 1:2. In other words 1 reward for 1 action, 1 reward for 2 actions and so on.
As I understand it, you C/T (clicker/treat)every time and very heavily at the start of the new behaviour. Once the dog has the idea, you can start phasing out the C/T by rewarding the dog at decreasing intervals leading towards C/T at totally random intervals (schedule?). Decrease your # of rewards very slowly and carefully, occasionally rewarding heavily and lots. The quality (smellier, tastier) of your treats also makes a huge difference to the dog. Extinction -as I think of it, is exactly what it means. The dog does not get any satisfaction from doing what it is doing so it stops doing it. Continuous probably means what you defined. I think that this is usually used at the beginning of learning something new, then gradually is changed to a random reward system. Reward does not necessarily have to be treats - it could be and is later something that the dog really wants ball, tug, retrieving something, chasing something, praise, etc. It could be something different with every dog.
From what I've read and found out personally, if you reward on a predetermined duration your dog will figure this out pretty quickly and start messing with you in order to get the treat quickly. For example at heel 1 reward every step, then every 2 steps, every 3 steps etc. The dog will figure this out and not give you want you are looking for. However if you say reward for the first step, the 3 steps, the 4 step, the 7 step (randomly) the dog will focus on the task at hand as it doesn't know when you are giving it what it wants. Obviously you have to be able to read your dog and understand where it is in its training and maturity and reward on this basis. Over time this C/T is decreased and something else is introduced and takes its place.
The strange part of all of this is that you don't put a name to the action until after the dog is doing it consistently. All the dog knows is that if it walks beside you in a certain place, it receives a treat.
There is a ton of info on this on the internet. Look for something that makes sense to you and use it. I suggest that when you do find a system stick with it unless you have a good reason to stop using it or switch to a different method.
Wayne
|
|
|
Post by hicntry on Feb 24, 2010 9:18:24 GMT -5
Wayne, I have talked to a lot of trainers on working boards, versatile dog boards and most other kinds of boards and notice that the passive methods used today donh't work well unless the handler is there handling the dog. Most trainers seem to have to crate their dogs when they leave the house for fear of property damage etc. Even harder to do with multiple dogs. Another scenario they don't seem to be able to overcome is when the dog is working without the handler at say 1 mile or more away in cattle country. How do you emply these methods to maintain total control of the dog so you know beyond a doubt when there are several dogs out together, they are not chasing someones calves.
|
|
|
Post by brandon on Feb 24, 2010 10:05:40 GMT -5
Wayne, the e-collar and force fetch have its place in bird work, however you may not need to have your dog forced to retieve if she has the drive to do so, i use the force fetch on dogs that have to go and retrieve every time say for a compition or in police work,however i have had other dogs that i help build there drive up and they would retrieve anything i asked. One thing to remember is we dont wont any birds left behind that are dead or crippled so forced retrieves can be a good bet you get your birds and dog back in one piece. The other thing is the cast away you will have some different challenges when teaching this with out force as well. this to can be done if you take your time and teach each small step and then proof the steps and you will see these things all come together, and if not you can always go to force, and with a terrier that might be what you need if she is like any dogs i have ever owned. lol
|
|
jax
Hunter/worker
Posts: 133
|
Post by jax on Feb 24, 2010 10:58:47 GMT -5
Jax, I think that I know what you are talking about regarding reward schedule. 1:1 1:2. In other words 1 reward for 1 action, 1 reward for 2 actions and so on. Extinction -as I think of it, is exactly what it means. The dog does not get any satisfaction from doing what it is doing so it stops doing it. Continuous probably means what you defined. I think that this is usually used at the beginning of learning something new, then gradually is changed to a random reward system. Reward does not necessarily have to be treats - it could be and is later Obviously you have to be able to read your dog and understand where it is in its training and maturity and reward on this basis. Over time this C/T is decreased and something else is introduced and takes its place. Wayne I guess you may have a handle on it. I asked Michael and he listened to my reading your post. His comments were... Click/Treat is a form of reward training...one of many. Reward schedules maintain the strength and value of the reward (from the dog's point of view) Once the dog understands that he can earn a payoff, he must then understand that there is also a consequence for noncompliance (Compulsion or punishment) Learning theory and stimulous control are the major tools to be understood before training any behavior Reward schedules give training structure which the dog intuitively understands and wants, it imparts a feeling of control in the dog. You mention the dog anticipating a reward delivery....that was the moment the training should have moved to a different reward schedule. Random training, or reward delivery, is a time waster and diminishes learning efficiency Too many trainers get sucked into thinking that only one type of reward is best, ie food. Rewards can and should be varied from the very earliest demonstrated understanding a dog shows. In doing so you maintain the value of all your rewards and maintain their desirability over distractions. There is a place for negatives, like e collars, in dog training. Dismiss them and you will discover giant holes in your attempt to finish yor dog. You don't mention in your post any reference to the dog hunting for the retrieve object. You have to connect your efforts training with a natural behavior or sequence of behaviors. Once the dog makes the connection, learning escalates from a demonstrated understanding/association of a stimulous directive and responsive behavior to problem solving. The dog's nose is his tool, make him use it. Relying only on visual retrieves as a training scenario is a time waster. He said that a buddy of his, Dave Kroyer, is putting out a new DVD on clicker training. Michael thinks highly of Dave and it may be worth your while to check it out.
|
|
|
Post by oksaradt on Feb 24, 2010 13:38:35 GMT -5
I think there is a lot of confusion of different training systems here. I know the clicker trainers would be having a fit reading the explanations. Any system that relies solely upon reward gambles that the dog associates that reward with actions ingrained with lots of repititions and disregards self-rewarding. The e-collar is a convenience that has been sold as a must-have tool. Koehler did quite well without it. Despite his bad rep, he had a lot of useful ideas and one of them was *cough* baby steps to the end result.
Nothing personaly towards you Jax, but your cousin tends to think higher of himself than most others do.
Regardless of training system, be it operant conditioning, compulsion, survival, yada yada yada....it all comes down to timing, consistency, skill building in the reward system, and a system of corrections. The corrections should be seen as just another part of the communication system between handler and dog. It is not about power. In a recently sighted utube post on fetching, the ear pinch is demonstrated by a handler with a nice touch. Just like with experimenting with clicker training without understanding the principles, applying an ear pinch incorrectly FOR THAT DOG can confuse rather than clarify. The message sent to the lab was "fetch the dummy or you get a pinch" and the trainer then states this is often the preliminary to using the e-collar. Dogs are sometimes described that they can't take "PRESSURE". This usually implies the handler doesn't know how to adjust the stimulus to that dog. Labs (in general) are easy to train because they tend to accomodate what ever rules the handler/trainer throws at them as long as the handler is consistent. To the handler's credit (in that utube) he followed up his ear pinch threat with verbal praise. Some handlers that are comfortable with compulsion, forget a follow-up reward. Such handlers are no better (or worse) than the clicker trainer that only uses positive reinforcement and no negative. You can run experiments with people (this can be a great party game) where you put three persons (the dogs) in your version of an isolation booth. The group decides some act that the "dog" must complete. The first person can only say "yes" to the "dog" if it works toward the end goal. The second person can only say "no" when the dog works away from the end goal. The third person gets to use "yes" and "no". Results tend to be consistent that the "dog" trained solely on negative reinforcement gets frustrated and may give up. The "dog" on positive reinforcement only may eventually get the goal completed with still some frustration. The dog that gets yes/no, hot/cold, etc. usually figures out the goal fairly quickly and everyone comes out happy.
I don't particularly like the e-collar any more than I do the clicker because it tends to only present one reinforcer. I realize a lot of people like the e-collar because it gives them contol when the dog is 0.5 miles away. Such control off-lead without an e-collar is only developed over gradually building distances of trust between the dog and the handler. I've yet to see a border collie on the hills of Scotland sporting an e-collar when they are herding sheep for the handler. For someone that has to send their dog off into the woods where they are out of sight of the handler, an e-collar is a dangerous temptation. What if the dog lingers working out a scent problem and the handler thinks they've been gone too long. The handler does a recall. The dog doesn't appear in the amount of time the handler expects, so they hit the e-collar and perform another stim. The dog could have been en route back to the handler and then gets corrected for doing so......What's a dog to do? For those venues that always have their dogs in sight such as retrievers and narcotics, I suppose an e-collar is quite the luxury.
How did we ever train dogs without them in the past? One just has to wonder, aye?
Wayne, what you demonstrated on your utubes was not a clicker system. You dd not mark the appropriate response but rewarded the end goal with lots of encouragement in between. Your system is just fine, but it's not clicker.
One version of a random reward system is utilized when performing a repetitive task multiple times and the reward becomes diminished in its attraction. Such an application would be documenting 200 graves in a forgotten cemetery. Give the dog a treat for each grave and it quickly fills up. Even ATF dogs only have to make 50 finds for their kibble per day. So, like compulsive gamblers you convince the dog over time that the jackpot is with the next pull of the one-arm bandit. I've successfully used this system with rewarding the first grave found, then then third, then the seventh, etc. I find the dog reaches a mental limit where it demands reward and you can use this to build drive, reward quickly and take it away, then go back to work. This is why I reward with a ball then trade-for-food. The dog gets a stress relief with the ball and then you quickly get it back to work with a food trade. There are dogs that work solely for verbal praise. I tend to see them limited to obedience and agility work where the work is as much fun for the dog as the reward. The dogs that do frisbee demonstrations are problem in their version of a dream job......course they are expected to give it up.
My current puppy has very high retrieve drive and I usually wear him out at night with about an hour of retrieve. Right now, he has a hard time deciding between the reward ball and the remains which he is also highly attracted to. In my mind, I'm the drug pusher as I'm giving him lots of free samples at night with the ball. Soon, that will stop and he'll only get to play retrieve for a human remains find. Usually, once that switch comes about the dog often doesn't want to give up the ball. I could pinch his ear and force him to release it with an "Out", but why diminish his reward with a correction. It's just as easy to slip him a piece of chicken jerky in trade for the ball. Having this system I can reward my dogs from six feet away to 40 yards away. Furthest find away from me for one of my dogs (in my sight) was about 200 yards. I asked the dog to wait and turned it into a commitment exercise because the dog is supposed to stay with the find until I recognize it. So, I ran as fast as I could till I could reward. Make me an e-collar that supplies a ball out of nothing as well as provide a correction and then I might consider them useful.
Dog's association time is 1-7 seconds. Association with an act is best marked within the first three seconds if not faster. Cats are around 1 second tops which is why a skilled clicker trainer can teach them tricks. Without an e-collar, long lines and throw-chains come in handy until you build the foundation for the dog for directionals. An e-collar is simply an invisible leash in skilled hands. In unskilled hands it is utter confusion and hell for the dog.
Jim
|
|
|
Post by ed on Feb 24, 2010 14:05:04 GMT -5
"it all comes down to timing," Don't need to say much more except 1)reward that is ( clicker, food praise, whatever) reinforces drive 2) correction( harsh voice, collar jerk,e collar) is the foundation of consistency
|
|
phil
Hunter/worker
Posts: 205
|
Post by phil on Feb 24, 2010 14:20:56 GMT -5
Wayne's world Wayne's world... If Clicker training is what you really want to do, and I think it would fit you and your dog to a tee. Wayne you really should get in touch with Corally Burmeister she be the Airedale Clicker meister... she and her disciples have done much in the area of canine behavior modification Not sure if she has ever made a gundog with it but you might be the first. There would be folks on this list that would know how to get you in touch with her..
Cheers
Hal
|
|
|
Post by maugh on Feb 24, 2010 15:19:12 GMT -5
www.clickertrain.comBut as Ed says, "it all comes down to timing." And with bad timing an intended reward (food, clicker, praise) becomes a punishment (a negative reinforcer, since it reinforces the wrong activity and shuts down the desired one) and conversely a poorly timed intended punishment (e-collar nick, leash pop, harsh voice) becomes a reward because it is associated with the wrong activity in the dog's mind and the original deed goes uncorrected (and if it was pleasurable to the dog, just doing it has reinforced it.) Timing is only learned by getting a coach. Regards, Maugh
|
|