Some in progress with collaborators. Mostly for later reference (often much later). See also here.

• Leave one out CV and Lasso
• Privacy and kernel density estimation
• Rademacher complexity like thing with empirical distribution of the data

Found this recipe on Epicurious. Their recommended cooking procedure is dumb. Meatballs must be browned! Very tasty, though barely enough food. Nearly finished off even though one roommate was a no show. Surely sufficient for 4 normal sized portions.

Meatballs:

• 2 pounds ground beef
• 1/3 cup minced onion
• 1/3 cup panko
• 1/4 cup chopped fresh cilantro
• 1 large egg, beaten to blend
• 2 garlic cloves, minced
• 1 tsp turmeric
• 1/2 tsp each, cayenne, cinnamon, nutmeg, grated ginger
• Salt and pepper

Stew:

• 2 chopped onions
• 4 garlic cloves, minced
• 2 cinnamon sticks
• 1 tsp turmeric
• 2 cups beef broth
• 1 14.5-ounce can diced tomatoes
• 1/4 cup golden raisins
• 2 cups 1/2-inch-thick carrot slices (cut on diagonal)
• 1/4 cup chopped fresh cilantro
• 1 package baby spinach leaves

Make 1.5 inch meatballs. Brown them nicely. Saute onions for 15 minutes on medium low. Add garlic and spices and saute 2 minutes. Add tomatoes, broth, raisins, carrots and cilantro. Submerge meatballs in the sauce. Simmer over low heat for 30 minutes. Add spinach and simmer five minutes more. Serve over couscous mixed with more cilantro, lemon juice, slivered almonds or pine nuts.

While the majority of this blog is dedicated to cryptic notes which serve only to remind me of my tasks for upcoming research meetings, I have decided to fill a void on the internet with writings on the subject of beef. In particular, James, Darren and I have just returned from Aaron Miller’s grass fed livestock farm with half of a delicious cow (butchered).  It is now safely stored in our basement in a 7 cubic foot chest freezer. We purchased 275 pounds of cow, which has now become ~150 lbs of deliciousness. Obviously, this will require the creativity and cooking skill my roommates and I possess in order to create tasty comestibles. In that spirit, I will attempt to post recipes and photos of many of the dishes we create. Of course 50+ pounds of ground beef will turn into numerous hamburgers, the recipes for which are obvious, but the more creative attempts (what to do with the rather large heart, etc.) as well as how we use all of that without inducing taste bud boredom will be posted here.

To do:

• Show the infinite order beta mixing converges to what we want
• Revise the text of the document (rates of convergence, clear up how to find the histogram)
• Investigate Yu’s results. What do we really need to estimate?
• Is there a less restrictive notion of mixing that gives us what we want?
• Chase the uncertainty through into the GEB
• Stability or Rademacher bounds for more complicated processes

1. Equation 15 in the report seems correct. It appears a judicious choice of ε and t should give you something like that, but how to make such a choice is not clear.
2. Examples of η-mixing processes are necessary. (Should be available from Leo’s other papers.)
3. A month or so ago, we discussed β-mixing for AR(p) models. There is still concern as to how this works for β(m) with m less than p. How is β-mixing for this scenario defined? Also, is it a problem if this is not well defined? To be clarified next week.
4. Along the lines discussed previously, is β-mixing necessary for the Mohri-Rostamizadeh type bound? It would be better if we could use this η-mixing definition. η-mixing also has a TV-distance form which would be amenable to the histogram type approach.
5. If AR models are an η-mixing process, bound the β-mixing estimate, see what happens.
6. See the paper on Levy processes by Tong and Zhang (linked on delicious).