Trained Parrot BlogParrot Wizard Online Parrot Toy StoreThe Parrot Forum

Continuous vs. Variable Ratio Reinforcement for Flight

Discuss topics associated with teaching birds to fly. Training parrots recall flight, target flying, and other flying exercises.

Re: Continuous vs. Variable Ratio Reinforcement for Flight

Postby Michael » Tue Mar 30, 2010 9:22 am

I continued expanding the reward ratio and tried getting into the VR15s and eventually VR20. My goal is basically to push it and discover how far I can go and then step back and stay within that threshold to get the most I can out of her. I don't think I'll ever be able to eliminate the treat as a reward for the recall but I think I will be able to give the treat so rarely and still have the bird recall that it would seem nearly like that.

Day 9:
60 recall attempts, 4 treats, 115g, several other tricks, motivation
3/5, misc seed treats

More treats for go to perch return than recall

50/60 1st attempt recall 83%
9/60 2nd attempt recall 15%
1/60 missed recall on 2nd attempt 2%

#1 early vr10:
8/10 1st attempt recall 80%
2/10 2nd attempt recall 20%

#2 mid session VR15:

12/15 1st attempt recall 80%
3/15 2nd attempt recall 20%

#3 mid session vr15:

12/15 1st attempt recall 80%
2/15 2nd attempt recall 13%
1/15 not recall within 2 calls 7%

#4 end of session vr20 (half distance):

18/20 1st attempt recall 90%
2/20 2nd attempt recall 10%
User avatar
Michael
Macaw
 
Gender: This parrot forum member is male
Posts: 6286
Location: New York
Number of Birds Owned: 3
Types of Birds Owned: Senegal Parrot, Cape Parrot, Green-Winged Macaw
Flight: Yes

Re: Continuous vs. Variable Ratio Reinforcement for Flight

Postby skeetersunconure » Wed May 12, 2010 8:11 pm

wow good job mike! you did alot of work haha. when i do training with skeeter i randomely give treats like he could go for 5 or 6 times without a treat just a click and sometimes he gets a random jackpot like of 2-3 seeds. :) btw how is outdooor harness flights coming along?
User avatar
skeetersunconure
Poicephalus
 
Gender: This parrot forum member is female
Posts: 386
Location: georgia
Number of Birds Owned: 1
Types of Birds Owned: sun conure
Flight: Yes

Re: Continuous vs. Variable Ratio Reinforcement for Flight

Postby Michael » Wed Nov 13, 2013 9:27 pm

I haven't recorded a formal variable ratio reinforcement schedule performance chart in a very long time but I have noted major improvement in my parrots' physical capabilities lately. They have become accustomed to flying greater distances outdoors as outdoor freeflight and thus have been performing substantially better indoors. They are at optimal healthy weights and in very good physical shape. They are both flying at their best capability so I pushed them to their limits to see how much they can fly and for how little.

The only treats used were Roudybush Medium Maintenance pellets. Both parrots flew for their entire dinner and received no additional supplement. Kili received half sized pellets while Truman received full size. They alternated flying recalls without a substantial break for rest for a session that lasted 60 minutes. Each flight was 50 feet from perch to me and then another 50 feet to return to the perch. Only recall flights were rewarded.

Here are the results of Weds November 13, 2013 for Kili (Female Senegal Parrot Age 5) and Truman (Male Cape Parrot Age 3):

Truman - Cape Parrot
80 recall flights, 23 treats, 305g, motivation 5/5, pellets

73/80 1st attempt recall 91%
7/80 recall not on first call 9%
VR3 used throughout

Total number of recalls - 80
Total number of flights - 160
Total distance flown - 8,000feet (1.5 miles)


Kili - Senegal Parrot
135 recall flights, 23 treats, 111g, motivation 5/5, pellets

115/135 1st attempt recall 85%
20/135 recall not on first call 15%
VR3 used throughout

Total number of recalls - 135
Total number of flights - 270
Total distance flown - 13,500feet (2.6 miles)


This is a record number of flights and distance flown in a single session for both parrots. Prior to this, the largest number of flights recorded by KIli was 72 and probably not more than 40 by Truman. Both have nearly doubled on their flying capabilities of a few years back.

Although Kili's stamina has improved, it appears that her recall reliability has remained the same. During the trials I performed 3 years ago, it appeared that she recalled on the first attempt with a rate of 85% and it was exactly such on this flying session. However, her reliability to come within 2 calls is closer to 95% and within 5 calls is 100%. Truman is less accustomed to flying on thinner variable ratio reinforcement schedules and it will be something to work further on.

Both parrots were tired by the end of the session but not completely worn out. Both had the capability to fly more but their crops were getting full. Recall response rate and eagerness were dwindling so I ended the session at the point I did. Kili consumed 12g of food and water. Truman consumed 25g of food and water. They both received the same number of pellets as rewards (coincidentally) but Kili flew on a thinner reinforcement schedule for more flights. Truman received twice the size in pellets and likewise gained twice the weight after the session. This appears to be a good comparison between the performance capabilities of both parrots at this stage. Keep in mind that Truman was never clipped and Kili was clipped once without ability to fledge.

I hope you find these flight training results interesting and challenge your parrot to fly as much as possible.
User avatar
Michael
Macaw
 
Gender: This parrot forum member is male
Posts: 6286
Location: New York
Number of Birds Owned: 3
Types of Birds Owned: Senegal Parrot, Cape Parrot, Green-Winged Macaw
Flight: Yes

Re: Continuous vs. Variable Ratio Reinforcement for Flight

Postby Wayne361 » Sat Nov 23, 2013 5:26 pm

Great reading. I also use vr rewarding and find I get a better flight response time from bird. I chalk this up to the bird knowing the reward is coming eventually and thus wanting to perform flights in quicker fashion as a means to the end. I also get the training benefits that Micheal mentioned i.e. longer training sessions, higher motivation, less treats, etc.
I also vary up treat type provided and find this increases motivation. I usually use pellets (broken up pieces) for main reward and offer a half sunflower seed and/or almond (small piece) in different ratios to main treat. I find a good ratio is about once every 5-8 rewards is "special reward" (seed/nut). Much like response time, by varying the treats with special rewards, my theory is that he knows a special treat is coming but doesnt know when. Thus he pushes for that special treat during training. This is all, of course, observatory but as such has assisted in Oscars training and maximizing training times/motivation.

Wayne
Wayne361
Conure
 
Gender: This parrot forum member is male
Posts: 220
Location: Oshawa, Ontario
Number of Birds Owned: 1
Types of Birds Owned: Senegal Parrot
Flight: Yes

Re: Continuous vs. Variable Ratio Reinforcement for Flight

Postby Michael » Sat Jan 25, 2014 10:27 pm

It was interesting to follow up with another variable ratio reinforcement schedule trial run today. I saw high motivation and had the time to fly the birds a lot and wanted to see how they would do on thinner reinforcement ratios. For the first 2/3 of the session, Kili & Truman flew thinner ratios then they have in the past. However, they were starting to get extremely fatigued and motivation was plummeting so I gave them a break by switching to continuous reinforcement for a little while. You have to realize that eating a pellet is also time consuming and provides greater rest than continuous back to back flying sessions. On a future trial, I would like to use a timer to guarantee a certain rest interval to take fatigue out of the VR equation as it pertains to food. I believe I could have had similar response on this session with less food reinforcement if it weren't for severe fatigue setting in (particularly for Truman who was panting at times).

Motivation was so high that the birds flew to their physical limits. The high pace of the training session got their body temperature to rise noticeably and for rapid breathing. The session lasted 45 minutes and aside from 2x 3 minute breaks, the birds flew continuously in alternating fashion. While one bird flew a recall, the other rested but had to be prepared to go as soon as the other returned.

Kili did not miss a single recall on the first call until the 31st flight recall, and not another until the 49th. After that point she had more missed flights I suspect due to fatigue rather than lack of motivation. Given a chance to rest she would start flying eagerly again. Both birds flew immediately on the first call during the 10 back to back continuous reinforcement trials near the end. Both continued to fly well on a VR schedule after the continuous reinforcement none the less.

I would say that the body condition of both birds is inferior to the condition they were in during the last trial. During the last trial both birds were receiving frequent outdoor flying sessions as well as indoor. Now they have been couped up in winter and only get the flying they can manage at home. Still, they both exceeded a mile of flying and both had flown over a quarter mile during the morning session. They reliably fly a mile indoors on most days in several sessions and play time. Their weights are also higher than they were on previous sessions but motivation was still excellent.

Here are the results of Saturday January 25, 2014 for Kili (Female Senegal Parrot Age 5) and Truman (Male Cape Parrot Age 3):

Truman - Cape Parrot
65 recall flights, 26 treats, 313g, motivation 4/5, pellets

59/65 1st attempt recall 91%
6/65 recall not on first call 9%
VR5 used for first 30
FR1 used for 31-40
VR3 used for 41-65

Total number of recalls - 65
Total number of flights - 130
Total distance flown - 6,500feet (1.2 miles)


Kili - Senegal Parrot
100 recall flights, 21 treats, 114g, motivation 5/5, pellets

90/100 1st attempt recall 90%
10/100 recall not on first call 10%
VR3 used throughout

Total number of recalls - 100
Total number of flights - 200
Total distance flown - 10,000feet (1.9 miles)


What is unbelievable is that the accuracy of response (coming on the first call) is pretty much identical across sessions. Truman has exactly 91% first recall attempt reliability like the previous session. Kili's response rate actually improved a bit and both maintain over 90%. If counting only the first 50 flights for both birds their response rates would be 96% for Kili and 92% for Truman. Their response rate begins to diminish as tiredness kicks in. In other words, for a few flights these birds are extremely reliable and most of the less reliable flights are the result of pushing them to fly more than usual with little resting time. Another interesting coincidence is that the number of pellet rewards received is similar to last time. This just happens to be the amount that they need to consume to make up the energy used for the flying and the rest for 12 hours of living.

Kili works out to be on an average VR5 for the entire session and Truman on VR2.5. This is fairly consistent with past trials. My next trial should include greater periods of mandatory rest to compare the role fatigue plays against food reward based motivation.
User avatar
Michael
Macaw
 
Gender: This parrot forum member is male
Posts: 6286
Location: New York
Number of Birds Owned: 3
Types of Birds Owned: Senegal Parrot, Cape Parrot, Green-Winged Macaw
Flight: Yes

Re: Continuous vs. Variable Ratio Reinforcement for Flight

Postby Michael » Tue Jan 28, 2014 9:27 pm

I decided to follow up on the previous session and try to rule out fatigue as a factor in the effectiveness of variable ratio reinforcement schedules to drive motivation. Yesterday I thought I was going to have good motivation so I attempted a flight session with additional 15 second breaks prior to each bird's flight. However, this turned out to be very distracting because they would want to fly sooner and then be distracted by when it was time to fly. Also motivation turned out poor and the session was terminated early.

Today I made another attempt but changed the system to include a 60 second break every 10 flights. Also I took care not to fly the birds to rapidly that they would become exhausted. I first performed a 60 flight session for the VR trial and then later had two more mini sessions which were not as stringent but I still totaled the figures to report here for the total flown in a 1.5 hour evening timespan.

Interestingly, reliability is even better when the birds are ensured to get sufficient rest. Occasional missed flights are most likely caused by some kind of distraction and not for lack of motivation. I can achieve very high reliability rates for flight recalls performed on VR5 and VR10 ratios. Kili did 60 flights for just 6 small pieces of pellet. That is over a mile of flying for what would amount to about 4 whole medium size pellets.

Motivation and tiredeness were intermittent in the less stringent follow up sessions. Reliability dropped off and I had to maintain more rewarding ratios to keep the birds on track. With a lot of effort milking the last bits of motivation, I drove Truman to surpass 100 flights for the first time and exceed 2 miles of flying in a single session.

Here are the results of Tuesday January 28, 2014 for Kili (Female Senegal Parrot Age 5) and Truman (Male Cape Parrot Age 3):

Truman - Cape Parrot
Part 1
60 recall flights, 12 treats, 304g, motivation 5/5, pellets

56/60 1st attempt recall 93%
4/60 recall not on first call 7%
VR5 used

Number of recalls - 60
Number of flights - 120
Distance flown - 6,000feet (1.1 miles)

Total with additional bonus session
100 recall flights, 27treats, 304g, motivation 5/5, pellets

89/100 1st attempt recall 89%
11/100 recall not on first call 11%
VR3.7 Overall

Total number of recalls - 100
Total number of flights - 200
Total distance flown - 12,000feet (2.3 miles)


Kili - Senegal Parrot
60 recall flights, 6 treats, 114g, motivation 5/5, pellets

57/60 1st attempt recall 95%
3/60 recall not on first call 5%
VR10 used throughout

Number of recalls - 60
Number of flights - 120
Distance flown - 6,000feet (1.1 miles)

Total with additional bonus session
100 recall flights, 13 treats, 114g, motivation 5/5, pellets

88/100 1st attempt recall 88%
12/100 recall not on first call 12%
VR7.7 Overall

Total number of recalls - 100
Total number of flights - 200
Total distance flown - 12,000feet (2.3 miles)
User avatar
Michael
Macaw
 
Gender: This parrot forum member is male
Posts: 6286
Location: New York
Number of Birds Owned: 3
Types of Birds Owned: Senegal Parrot, Cape Parrot, Green-Winged Macaw
Flight: Yes

Re: Continuous vs. Variable Ratio Reinforcement for Flight

Postby nerdybird » Thu Mar 05, 2015 3:49 pm

This is fantasic info for recall training.

I've been recalling my parrotlet for a few months now, and this post is very inspiring, one because I don't feel as bad about her not being at 100%, and two, I am motivated to track our training data to have accurate info about her progress over time.

She was clipped when I first got her in August 2014, but the last few days i've noticed her really doing some great flying on her own. I checked out her wings and 3 of her outer flight feathers have grown back in. It is amazing what she can do now! She will come looking for me now, and she can fly across the apartment and back; previously she seemed to barely make it across once and if she couldn't land she would just crash. Now she can make adjustments midair, and she is noticeably faster. She has been flying of her own agency a lot also, and doing some impressive aerial maneuvers.

I will definitely start keeping track of her response rate, and start using variable rewards. I thought that her lack of attention was due to her being a smaller bird; I seem to remember reading that they can have shorter attention spans. Mostly she will be doing well, and then she will go back to her perch and suddenly start preening and stop paying attention to me :p
nerdybird
Lovebird
 
Gender: This parrot forum member is female
Posts: 33
Location: Seattle
Number of Birds Owned: 2
Types of Birds Owned: parrotlet, cockatiel
Flight: Yes

Previous

Return to Recall Flight & Flighted Tricks

Who is online

Users browsing this forum: No registered users and 3 guests

Parrot ForumArticles IndexTraining Step UpParrot Training BlogPoicephalus Parrot InformationParrot Wizard Store