Samuel had a dry night last night! It all started when he asked for his “six o-clock”. Oh, maybe we need to start further back: We have a rule in our house that Samuel can’t get up until 6:00. Usually that involves trying to convince the 3 year old that all the clocks in the house say 5:00 (which isn’t 6:00) and that the sun really isn’t actually up yet. So now every clock, timer and circular thing with numbers on it, is a “six o-clock”. His “six o-clock” is an egg timer that we use when he goes to his room, or needs to take turns, or needs a reminder for those annoying potty brakes. He tends to ask for it at night when he goes to bed, and since it is only a 60 minute timer, we take it out of his room when he has fallen asleep.
But we had company last night and in all the fun I forgot that it was there. So at 8:30 his timer went off. Loudly. And up jumped Samuel, telling us that his “6 o-clock” said that it was morning. And since both Mommy and Daddy were awake, he must need to be awake too. So I told him that his timer was just telling him that he needed to potty, and then he would go back to bed. Because 8:30 pm is deep in the middle of the night for a small Samuel, he willingly did everything that I told him. And because he was so wonderful about it all I allowed him to keep his “6 o-clock” - running down, of course.
This time when he fell asleep, I took his timer. And did so again, at 1:30, because my Samuel could tell that his “6 o-clock” was gone and woke me up so I would put it back in his room. I didn’t do it at 4:30 because at that point Samuel got in bed with us and the timer isn’t allowed while I’m sleeping. Each time we made sure his diaper was dry before he got back in bed.
And so, we had a very excited Samuel who knew he had achieved something fabulous. And is still, at nap time, talking about what he did. I only hope that his mother can keep up with his awesomeness.
Wednesday, February 29, 2012
Tuesday, March 22, 2011
Healthy!
I've been rejoicing this past couple of days and I just want to let all of you know: My back doesn't hurt, my ribs are in place, I have energy, and I'm not sick! I thank God for a body that is healthy and is working.
As you might know I've been struggling with health issues for quite some time and then this winter I feel like I've been sick for several months straight. The aftermath of several different colds brought my body down to the lows I felt when I had Chronic Fatigue. It was stressful, depressing, and really hard for me to deal with. But these last couple of days I've been dancing as I go about my duties. The affirmation that I have a body that can do remarkable things has filled my head with so much gratitude that I sometimes feel like crying when I do dishes, or vacuum, or play with Samuel - because I can do those things. I am so grateful that I am currently living and experiencing my life with a mind that is free of pain and fatigue.
I am feeling so hugely blessed, in part, because I know that God was with me when I was sick. When I contemplated Him, I was able to see my life as productive and meaningful. I was able to enjoy what I had and rejoice in it. My expanded blessings now make me look toward the future and the greater capacity to enjoy the love of my family, the wonders of the world, and the perfect health that God will give to me in my resurrected body. I want to be worthy of such blessings and I want to have the kind of mind that will appreciate all that He will make of me.
Suffering and the task of trying to be cheerful while in suffering has taught me that happiness and contentment are states of mind that we have the ability to enter at will. I have learned that when I feel love I cannot be afraid, or furious, or despairing - but rather, my heart is full of the wonders of my life, the gratitude that I have loving family, the precious body God has given me. And I have learned that I can develop a mindset that will allow me to experience heaven to its fullest glorious perfection. And not take any of it for granted.
I love my healthy body and I can only imagine how wonderful it would be to be free of the imperfections still rampant in my mortal body. It has become another motivation to strive to be worthy - to improve myself as much as I may so that I can see and experience the future miracles God will place in my life.
As you might know I've been struggling with health issues for quite some time and then this winter I feel like I've been sick for several months straight. The aftermath of several different colds brought my body down to the lows I felt when I had Chronic Fatigue. It was stressful, depressing, and really hard for me to deal with. But these last couple of days I've been dancing as I go about my duties. The affirmation that I have a body that can do remarkable things has filled my head with so much gratitude that I sometimes feel like crying when I do dishes, or vacuum, or play with Samuel - because I can do those things. I am so grateful that I am currently living and experiencing my life with a mind that is free of pain and fatigue.
I am feeling so hugely blessed, in part, because I know that God was with me when I was sick. When I contemplated Him, I was able to see my life as productive and meaningful. I was able to enjoy what I had and rejoice in it. My expanded blessings now make me look toward the future and the greater capacity to enjoy the love of my family, the wonders of the world, and the perfect health that God will give to me in my resurrected body. I want to be worthy of such blessings and I want to have the kind of mind that will appreciate all that He will make of me.
Suffering and the task of trying to be cheerful while in suffering has taught me that happiness and contentment are states of mind that we have the ability to enter at will. I have learned that when I feel love I cannot be afraid, or furious, or despairing - but rather, my heart is full of the wonders of my life, the gratitude that I have loving family, the precious body God has given me. And I have learned that I can develop a mindset that will allow me to experience heaven to its fullest glorious perfection. And not take any of it for granted.
I love my healthy body and I can only imagine how wonderful it would be to be free of the imperfections still rampant in my mortal body. It has become another motivation to strive to be worthy - to improve myself as much as I may so that I can see and experience the future miracles God will place in my life.
Wednesday, February 17, 2010
James Carroll's PhD Dissertation Defense
Title: "A Bayesian Decision Theoretical Approach to Supervised Learning,
Selective Sampling, and Empirical Function Optimization"
Thursday, February 25th at 9:30 am
B092 JFSB
Advisor: Dr. Kevin Seppi
*All are encouraged to attend!*
ABSTRACT:
Today’s computers are capable of performing many tasks that previously only people could perform. This is partly due to recent techniques that have allowed computer systems to “learn” from experience. Machine Learning (ML) is the field of computer science that attempts to understand what it means to learn, and how such learning can be performed.
One major missing ingredient in the study of Machine Learning is a cohesive theory and framework in which a broad range of questions can be analyzed and studied. An effective “theory” of machine learning should be able to answer questions in areas such as: learnability, bias, overfit, no-free-lunch (NFL), meta learning, transfer learning, selective sampling, ensembles, and sample complexity. Other important issues such as the connection between supervised learning and other fields such as un-supervised learning, semi-supervised learning, function optimization, compression, and information theory should also be handled by such a model.
Several paradigms or theories that attempt to formalize the learning process have been proposed including: PAC learnability, VC dimensionality, and the Extended Bayesian Framework (EBF). To date, none of the proposed frameworks have been able to answer all of these questions in a satisfactory manner. For example, in some cases it is possible to use VC dimension or PAC learnability to show that some problems are learnable and to place theoretical upper bounds on their sample complexity (the number of examples required to learn a given problem). However, in practice, such bounds are seldom if ever used, because they vastly overestimate the actual number of examples required. Furthermore, PAC learnability and VC dimensionality do not provide a framework in which the other questions can be easily answered. On the other hand, the EBF has been used to prove the NFL theorems, but is not useful for examining questions of learnability or sample complexity. To date, no single theory has been proposed which can deal with all of the above areas.
We propose a Unified Bayesian Decision Theoretic Model (UBDTM). The UBDTM is a unified model of un-supervised learning, supervised learning, semi-supervised learning, and empirical function optimization. This model expresses the dependencies and relationships between observed feature vectors, observed labels, unobserved (test and validation) feature vectors, unobserved labels, the function that maps them, and the location of the extremum (or of multiple extremum locations) of the function, and the value at that location. There are many advantages to thinking about machine learning and function optimization in this way.
It will be beyond the scope of a single dissertation to adequately address all of the questions issues in terms of the UBDTM, however, we believe that this framework will provide a sound theoretical framework in which all of the above issues can be eventually addressed. This dissertation will focus on addressing a few specific questions in the context of the UBDTM. Specifically, we will show that the UBDTM provides another way of thinking about the bias and No-Free-Lunch problems. We will show that UBDTM illustrates the importance of explicit utility functions in Machine Learning for decision making. We will show that given some utility assumptions a priori distinctions between learning algorithms are possible. We will show that existing NFL results do not actually show the ”futility of bias free learning“ and we will propose a new approach to showing the futility of bias free learning. We will extend the UBDTM to deal with active learning and selective sampling, thereby creating a ”theory“ of selective sampling, including an active learning NFL result. We will show that this approach can lead to useful application with a Bayesian training procedure for the CMAC neural network topology, and give an application to active learning by modeling the annotated corpus creation process. We will then give a theory of empirical function optimization, including giving the empirical function optimization NFL result in our model. Finally, we will provide a concrete example of using our model to solve an empirical function optimization problem. We will demonstrate the connections between active learning and empirical function optimization, and show that these connections flow from their shared model (the UBDTM). Thus, one advantage of the UBDTM is that it makes the connection between supervised learning and function optimization clear. Although we believe that the remaining questions of meta learning, transfer learning, learnability, and sample complexity can also be addressed in terms of the UBDTM we will leave these problems for future work.
The performance of these techniques will be demonstrated on several synthetic problems and on real world problems including a tagging problem which is part of a larger joint project to build an annotated Syriac corpus with the BYU Center for the Preservation of Ancient Religious Texts (CPART). Thus, tools derived from our model will be used to build a publicly available corpus of tagged Syriac tests.
Intellectual Merit: The UBDTM has the potential to advance our general understanding of the supervised learning and empirical function optimization problems, and various proposed solutions to them, by better understanding the underlying processes involved.
Broader Impact: Supervised Machine Learning problems are important in a wide variety of problems of relevance in a wide variety of fields, for example: linguistics (tagging natural language text); robotics (machine vision); homeland security (face recognition, machine translation); and medicine (disease diagnosis). Function optimization is one of the most common computational problems encountered in all of computer science. It is hoped that a better understanding of the statistical theory involved in supervised learning and optimization will lead to better algorithms and algorithm analysis. Testing will involve cross department research involving individuals from Computer Science, Linguistics, and CPART.
Keywords: Machine Learning, Supervised Learning, Bayes, Bayes Law, Bayesian Learning, Decision Theory, Utility Theory, UBM, UBDTM, Learning Framework, No-Free-Lunch, Bias, Inductive Bias.
Selective Sampling, and Empirical Function Optimization"
Thursday, February 25th at 9:30 am
B092 JFSB
Advisor: Dr. Kevin Seppi
*All are encouraged to attend!*
ABSTRACT:
Today’s computers are capable of performing many tasks that previously only people could perform. This is partly due to recent techniques that have allowed computer systems to “learn” from experience. Machine Learning (ML) is the field of computer science that attempts to understand what it means to learn, and how such learning can be performed.
One major missing ingredient in the study of Machine Learning is a cohesive theory and framework in which a broad range of questions can be analyzed and studied. An effective “theory” of machine learning should be able to answer questions in areas such as: learnability, bias, overfit, no-free-lunch (NFL), meta learning, transfer learning, selective sampling, ensembles, and sample complexity. Other important issues such as the connection between supervised learning and other fields such as un-supervised learning, semi-supervised learning, function optimization, compression, and information theory should also be handled by such a model.
Several paradigms or theories that attempt to formalize the learning process have been proposed including: PAC learnability, VC dimensionality, and the Extended Bayesian Framework (EBF). To date, none of the proposed frameworks have been able to answer all of these questions in a satisfactory manner. For example, in some cases it is possible to use VC dimension or PAC learnability to show that some problems are learnable and to place theoretical upper bounds on their sample complexity (the number of examples required to learn a given problem). However, in practice, such bounds are seldom if ever used, because they vastly overestimate the actual number of examples required. Furthermore, PAC learnability and VC dimensionality do not provide a framework in which the other questions can be easily answered. On the other hand, the EBF has been used to prove the NFL theorems, but is not useful for examining questions of learnability or sample complexity. To date, no single theory has been proposed which can deal with all of the above areas.
We propose a Unified Bayesian Decision Theoretic Model (UBDTM). The UBDTM is a unified model of un-supervised learning, supervised learning, semi-supervised learning, and empirical function optimization. This model expresses the dependencies and relationships between observed feature vectors, observed labels, unobserved (test and validation) feature vectors, unobserved labels, the function that maps them, and the location of the extremum (or of multiple extremum locations) of the function, and the value at that location. There are many advantages to thinking about machine learning and function optimization in this way.
It will be beyond the scope of a single dissertation to adequately address all of the questions issues in terms of the UBDTM, however, we believe that this framework will provide a sound theoretical framework in which all of the above issues can be eventually addressed. This dissertation will focus on addressing a few specific questions in the context of the UBDTM. Specifically, we will show that the UBDTM provides another way of thinking about the bias and No-Free-Lunch problems. We will show that UBDTM illustrates the importance of explicit utility functions in Machine Learning for decision making. We will show that given some utility assumptions a priori distinctions between learning algorithms are possible. We will show that existing NFL results do not actually show the ”futility of bias free learning“ and we will propose a new approach to showing the futility of bias free learning. We will extend the UBDTM to deal with active learning and selective sampling, thereby creating a ”theory“ of selective sampling, including an active learning NFL result. We will show that this approach can lead to useful application with a Bayesian training procedure for the CMAC neural network topology, and give an application to active learning by modeling the annotated corpus creation process. We will then give a theory of empirical function optimization, including giving the empirical function optimization NFL result in our model. Finally, we will provide a concrete example of using our model to solve an empirical function optimization problem. We will demonstrate the connections between active learning and empirical function optimization, and show that these connections flow from their shared model (the UBDTM). Thus, one advantage of the UBDTM is that it makes the connection between supervised learning and function optimization clear. Although we believe that the remaining questions of meta learning, transfer learning, learnability, and sample complexity can also be addressed in terms of the UBDTM we will leave these problems for future work.
The performance of these techniques will be demonstrated on several synthetic problems and on real world problems including a tagging problem which is part of a larger joint project to build an annotated Syriac corpus with the BYU Center for the Preservation of Ancient Religious Texts (CPART). Thus, tools derived from our model will be used to build a publicly available corpus of tagged Syriac tests.
Intellectual Merit: The UBDTM has the potential to advance our general understanding of the supervised learning and empirical function optimization problems, and various proposed solutions to them, by better understanding the underlying processes involved.
Broader Impact: Supervised Machine Learning problems are important in a wide variety of problems of relevance in a wide variety of fields, for example: linguistics (tagging natural language text); robotics (machine vision); homeland security (face recognition, machine translation); and medicine (disease diagnosis). Function optimization is one of the most common computational problems encountered in all of computer science. It is hoped that a better understanding of the statistical theory involved in supervised learning and optimization will lead to better algorithms and algorithm analysis. Testing will involve cross department research involving individuals from Computer Science, Linguistics, and CPART.
Keywords: Machine Learning, Supervised Learning, Bayes, Bayes Law, Bayesian Learning, Decision Theory, Utility Theory, UBM, UBDTM, Learning Framework, No-Free-Lunch, Bias, Inductive Bias.
Wednesday, December 16, 2009
Laughing.
Have you ever tried to do something coordinated while laughing? Laughing tends to make the task nearly impossible. Poor Samuel has figured this out...or I guess I should say, his mother has figured this out.
You know how you teach kids how to walk: You set them a couple steps away from you and then make them come to you. Well, Samuel thinks it is soooooo funny that he laughs so hard it turns him drunk. He wobbles, he tips and his legs just won't hold him up, and all the while he is giggling like the silliest baby in the world. Which then makes you laugh, which in turn makes Samuel laugh so hard that you have to stop just so you don't suffocate your baby while trying to get him to walk. Although, I will say it is beyond cute to see his floppy, shivering frame convulsing in laughter while you futilely try to get him to stand back up. And it only makes it worse when James tries to help. We all end up laughing so hard we can't do anything.
I see a parallel here: When you know that someone is going to catch you, you can enjoy the fall. God is always there holding his hands out in case we need him, and if we can remember that we can enjoy learning how to walk through this life.
Lately Samuel has started taking a couple of steps away from the furniture, before collapsing. And if he sees me noticing, he has to come over and drape himself over my knees to finish his giggle. He can actually walk those few steps by himself. So he is learning how to walk! Eventually he will start to walk and I hope that we can then find something else to giggle over - because enjoying this world with my sweet family is the most amazing thing I have ever done.
Tuesday, December 8, 2009
Dog Food
You have to assume that dog food is nutritious. Dogs eat it their whole lives and get every nutritional need from it. So, nutritious it is. However, it is still pretty gross to find Samuel eating it. It's bad enough that he wants to eat rawhide bones, those are easy enough to hide from him. But the dog has to eat some time. I kind of wish that Lylah would be more protective of her bones and food. It would make my job easier. But she has taken to depositing her bones on Samuel's lap and opening the door to her room so that Samuel can get in there to eat her food. Sigh. It's a no win situation. Either Lylah protects her food and thinks of herself as boss over Samuel, or Samuel is the boss and he gets whatever he wants. I've tried putting the food in a place where Samuel can't reach it, but then Lylah can't reach it either. My organizational skills just aren't that awesome. I guess I just hope that Samuel will get it out of his system before he is 6 and going over to his friends house to see if the different colored dog biscuits really taste different. And I guess if dog food is that nutritious, it really isn't hurting him...And I suppose that he is learning how to eat hard, crunchy and highly dissolvable foods...
Tuesday, November 17, 2009
Thursday, April 16, 2009
Obituary, Gwen Lenore Carroll (Mecham) 1936-2009
Gwen Lenore Mecham Carroll, a loving wife, mother, grandmother, daughter, sister and cherished friend, lost a courageous and a long-lasting struggle with breast cancer and passed away peacefully on April 15, 2009 at LDS hospital in Salt Lake City, Utah. She was born June 17, 1936 in Springville, Utah, the oldest child of Emil Lemond and Erma Fullmer Mecham. She attended school in the Springville and Provo areas, graduating from Provo High School in 1954. She worked at the Social Security Office and the BYU Employment Office to support her University experience. Gwen interrupted her BYU studies to serve an LDS mission in South Africa from 1959 to 1961. Upon returning from her mission, she began working as a flight attendant for United Airlines. She worked as a flight attendant for over 25 years, and was based in Chicago, San Francisco, and Denver. It was while she was living in the San Francisco area that she met and married the love of her life, Jimmy Lee Carroll, on March 17, 1964. Their marriage was solemnized in the Manti Temple in 1980.Gwen’s greatest joy in life came from being a mother to her son, James Lamond Carroll. She was also a wonderful step mother to Terry Carroll and Linda Strebeck. She was a master gardener, and year after year she produced yards that were some of the most beautiful in the valley. She also enjoyed dancing, playing the piano and singing with her ward choirs. Gwen loved history, genealogy, family history work, and reading and learning about her ancestor’s lives. Gwen was a life-long member of the LDS church in which she served diligently in numerous callings. She loved and was loved by her family, her extended family, her church associates, and a myriad of wonderful friends. She was indeed one of God’s choicest daughters.
Gwen is survived by her husband of 45 years, Jimmy L. Carroll, her son, James L. (Heidi) Carroll, her step-children Terry Lee (Vi) Carroll, and Linda Joyce (Alan) Strebeck, one grandchild and fifteen step-grandchildren and great-grandchildren. She is also survived by her four sisters Shirlene (Ron) Day, Karen (Rod) Marshall, Kay (Jim) Graff, Elaine Butler, and numerous nieces, nephews and cousins. She was preceded in death by her parents, Emil Lemond Mecham, Erma Fullmer Mecham Nelson, her step-father, William H. Nelson and a brother, Emil Lemond Mecham.
A viewing will be held on Sunday April 19th from 6 until 8 pm at Berg Mortuary, 185 E. Center Street Provo Utah. Funeral services will be held on Monday April 20st at 11 am, at the LDS Church, at 10494 N. 4720 W, Highland Utah, with a viewing one hour prior to services. Internment will be in the Springville Evergreen Cemetery.
Subscribe to:
Posts (Atom)
