Tweetovi

Blokirali ste korisnika/cu @ShaneBarratt

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @ShaneBarratt

  1. Prikvačeni tweet
    19. ruj 2014.

    Your entire life has led up to the moment you're reading this tweet.

    Poništi
  2. proslijedio/la je Tweet

    Shane Barratt, Guillermo Angeris, Stephen Boyd : Automatic Repair of Convex Optimization Problems

    Poništi
  3. proslijedio/la je Tweet

    Malayandi Palan, Shane Barratt, Alex McCauley, Dorsa Sadigh, Vikas Sindhwani, Stephen Boyd : Fitting a Linear Control Policy to Demonstrations with a Kalman Constraint

    Poništi
  4. 13. sij

    California is spending .1% of its budget on transportation... What a joke.

    Poništi
  5. 24. pro 2019.

    Code for all of the examples is available online: .

    Prikaži ovu nit
    Poništi
  6. 24. pro 2019.

    We provide examples in portfolio optimization ("Tuning a Markowitz policy to maximize utility"), vehicle control ("Tuning a vehicle controller to track curved paths"), and supply-chain management (" Tuning a supply chain policy to maximize profit"). Check them out in the paper!

    Prikaži ovu nit
    Poništi
  7. 24. pro 2019.

    Our (simple) method is able to reach the performance of a (sophisticated) method based on LMIs, introduced by Wang & Boyd ().

    Prikaži ovu nit
    Poništi
  8. 24. pro 2019.

    Next we apply our method to a box-constrained LQR problem, which has no known exact solution. We use the same COCP as the LQR problem except we add the constraint that the input is in a box.

    Prikaži ovu nit
    Poništi
  9. 24. pro 2019.

    As for numerical examples, we start by applying our method to the classical LQR problem, where the parameters are the coefficients of an approximate (quadratic) value function. We see that our method is able to recover a policy with the same cost as the LQR solution.

    Prikaži ovu nit
    Poništi
  10. 24. pro 2019.

    Our method relies on recently developed methods that can efficiently evaluate the derivative of the solution of a convex optimization problem with respect to its parameters, namely cvxpylayers ().

    Prikaži ovu nit
    Poništi
  11. 24. pro 2019.

    Tuning the parameters in these policies is often done by hand, or by grid search. In this paper we propose a method to automate this process, by adjusting the parameters using an approximate gradient of a performance metric with respect to the parameters.

    Prikaži ovu nit
    Poništi
  12. 24. pro 2019.

    Many control policies used in practice are in fact COCPs. Some examples include the linear quadratic regulator (LQR), convex model predictive control (MPC), and convex approximate dynamic programming (ADP) or control-Lyapunov policies.

    Prikaži ovu nit
    Poništi
  13. 24. pro 2019.

    We consider the problem of tuning convex optimization control policies (COCPs), which are control policies that compute the input or action by solving a convex optimization problem that depends on the current state and some parameters.

    Prikaži ovu nit
    Poništi
  14. 24. pro 2019.

    New paper - Learning Convex Optimization Control Policies, w/ , , and Stephen Boyd. Paper:

    Prikaži ovu nit
    Poništi
  15. 5. stu 2019.

    Had a great time interning at last summer. Check out this blog post describing my and a few other intern’s experience there!

    Poništi
  16. 4. stu 2019.

    Bay Area public transportation in one photo

    Poništi
  17. proslijedio/la je Tweet

    Our imaginations are being held hostage by the car.

    Poništi
  18. 29. lis 2019.

    To get a lower bound, we first convert the problem into a mixed-integer convex program using the perspective formulation () and relax the integral constraint to get a lower bound. (See the paper for more details.) Here is an example of the lower bound.

    Prikaži ovu nit
    Poništi
  19. 29. lis 2019.

    Our algorithm requires solving roughly 5-20 convex optimization problems, and we have implemented it as a CVXPY extension, making it easy to (approximately) solve such problems

    Prikaži ovu nit
    Poništi
  20. 29. lis 2019.

    And clipped control, where, for example, the cost encourages us to be in one of the two lanes

    Prikaži ovu nit
    Poništi
  21. 29. lis 2019.

    Applications include clipped empirical risk minimization, for example, clipped regression

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·