when i was on an AI kick, I had a spider bot. started out using negative reinforcement to train it, but it would still find content i did not like. the key was to train it with positive reinforcement, and it ended up getting lost in good content. now i use that same technique for my diet.

instead of concentrating on the foods i cant have, i just think about all the good foods that i should eat. there are enough good foods, that i really dont miss the processed crap. so i really like the book The 150 Healthiest Foods on Earth along with the website whfoods.com. along the way, i've been having to learn how to cook. as Andy suggested, i turn to Alton Brown and Good Eats. i like his series because each episode generally concentrates on a single type of food. alot of times, i'll decide to try one of the healthiest foods from a source above, and then i'll see if there is a Good Eats episode dealing with it. of course i just skip the desert episodes. there really needs to be more 'healthy' cooking shows, something like these Cooking with Christina episodes on bodybuilding.com. speaking of bodybuilding.com, i started out reading there nutrition articles to figure out the macro-nutrient percentages i should be aiming for. it also has an endless supply of bodybuilder diets that i can attempt to figure out which one works best for my body type. so i've really had to relearn what food is. this has required alot of experimentation. so i make it a point to get at least 1 new food every time i go to the grocery store. this has helped me to rediscover alot of foods and find many new ones. there's been a couple times where i've taken an unlabeled veggie up to the grocery store checkout line ... and nobody even knew what it was (i.e. parsnips).

as far as organics, i saw the Penn & Teller Bullshit episode. so i dont really bother with organic veggies or fruits. i'll sometimes get organics just because the package might be smaller. otherwise, i might get the organic variety of a processed food. granted, i dont eat many processed foods to begin with ... but the organic variety generally wont contain unnecessary additives like high fructose corn syrup. about the only thing i regularly get that is organic, is organic decaf coffee beans.

for the last 4 months, my bodyweight kept bouncing between 196 and 201 lbs; although the goal had been to get below 195 lbs. so i backpedaled and started experimenting with the diet. i tried a seriously low carb approach, but that destroyed my energy levels and ability to think. then i tried just getting starchy carbs before and after a workout. the problem with this approach is that i never seemed to wake up in the morning. so i found out that i require starchy carbs in the morning (to get started) and then before and after workout. that worked for about a week, and then my energy levels would start dipping.

what seems to be working now is to have a mandatory 'cheat' day. for 5 days, i eat 1500-1750 calories per day. 40 grams fat, around 125 grams carbs, and over 200 grams protein. these days are mostly lean meats, fibrous veggies, some starchy veggies and some grain. i dont get any fruit or dairy on these days ... so there is very little sugar. on the 6th day, i 'cheat', and get about 2500 calories. i change things up by eating over 250 grams of protein, around 300 grams of carbs, and only incidental fats (less than 20 grams). the additional carbs come from more grains and starchy veggies, plus lots of fruit and dairy (tons of sugar). this cheat day seems to keep my metabolism from becoming too efficient. it also keeps my energy levels from dipping too low.

on the new diet, my bodyweight just dropped to 194 lbs. this is significant because it makes a total loss of 50 lbs. for bodyfat calipers, my chest and thighs still measure around 4-6. my belly had been measuring 15, but now its down to around 10. for measurements, my waist dropped from 36 to 35.5. belly (measured at belly button) dropped from 36 to 35. love handles dropped from 37 to 36. so this is the first time my belly measurement dropped below my waist measurement. visibly, i still have some fat covering my lower abs. the love handles are smaller but starting to shape up. the new stubborn fat seems to be some lower back fat.

the new goal is to drop below 190 lbs. i'm going to keep the diet as-is until my energy levels start dropping. then i might switch from 5 days on 1 cheat, to 4 days on 1 cheat.

/dicResKit is a simple article providing an overview of the Dictation Resource Kit (DRK), which can be used to generate language models for speech recognition and speech synthesis.

integrated DRK grammars into /eva. first, the code works on both Vista and Windows 7. i'm still trying to figure out why the previous n-gram loading code worked on Vista, but not Windows 7. second, this allows me to automate generation of the grammars nightly on a deployment machine. the only requirement is installing the DRK to the deployed machine, which is much easier to swallow than installing Visual Studio and Speech Server 2007 to get the Conversational Grammar Builder installed. that said, i'd still rather have a simple executable that i could hand a corpus text file and get back a compiled n-gram (.cfg).

i've only done minimal runtime testing, but so far, the results from using an n-gram with AppendRuleReference seem to be better than using a DRK language model with AppendDictation(topic). that's strictly from looking at the SemanticValues returned from Recognition. second, i'd also have to rework my search logic. the problem is i normalize the music library in my own code and store the original values and normalized values to a dataset. but then the DRK can normalize the values again, so i'd have to remap that normalized value back to the dataset to get the search logic to find more matches. e.g. my logic currently normalizes '808 state' as it is, but the DRK will normalize to 'eight zero eight state'. for now, i'm just skipping the step of using the DRKs normalizer.

so i've still got problems. the n-gram seems to give better results but there isnt a way to automate generating grammars and the code isnt working on Windows 7. vs the DRK language models, which dont seem to perform as well recognition-wise for this task, but they can be generated automatically on a deployment machine and the code is working on Windows 7. er, um ... so i'm stuck with using the DRK for now.