Atendendo a algumas respostas apreciativas e inúmeras ignorativas, aqui está......

30 junho 2017

Relatório para a USP

Acabei de enviar dados para um relatório que a USP pediu. Os números da secretaria estão errados mas o navio está vazio, o relatório está preenchido, e eu não me chamo Honestinfus.

O relatório será usado com a finalidade de encher a urna para a distribuição de cargos. Só lendo nosso maior escritor para entender.

01 junho 2017

Inverse Laplace transform with fractional exponents

The inverse Laplace transform of


with a > 0, is

e^{-t} t^{-1 + a}/Gamma(a).

I'm still trying to understand what this means, but meanwhile here's a picture. Note that it tends to an impulse when a goes to zero. When a = 1, we see the familiar exponential decay. For larger values, the inverse transform starts at 0 and reaches a maximum before decaying exponentially.

(I don't know how to make sense of the unstable case 1/(1-s)^a, so perhaps all these calculations don't help understand anything.)

30 maio 2017

Machine Learning

Faz algum tempo acho que seria interessante oferecer uma matéria sobre machine learning para os alunos do PTC da engenharia elétrica da Poli. Acho que seria especialmente útil para a turma de Controle, e também de Teleco.

O processo burocrático para oferecer uma matéria nova é demorado demais para o ritmo da tecnologia. No 2o semestre de 2017 vou oferecer "Tópicos de Controle Avançado". Estou considerando ignorar a ementa e escolher machine learning como tópico principal. Se houver tempo e interesse, investimos algum tempo nos tópicos antigos também. É um sacrifício de uma matéria  interessante, divertida, e para mim pouco exigente. Mas para os alunos me parece um bom caminho. Se não for agora quando, e se não eu quem?

Só tem um pequeno problema: não sei nada sobre machine learning. Detalhe pouco relevante. Diz-se que a única forma de aprender um assunto é ensinando. Então sugiro que os alunos me ensinem. Quando tiverem terminado de ensinar, terão aprendido. Faz sentido - o objetivo é da escola é os alunos aprenderem, não o professor.

Motivação: dizem por aí que data science é o emprego do século 21. Diversas universidades mais ágeis administrativamente do que a USP têm criado programas em "data science" ou orientados para "big data". Machine learning são os algoritmos centrais, e as ferramentas são coisas bastante conhecidas pelos alunos do PTC. Quem estudar um pouco sai com 2 voltas de vantagem para trabalhar no assunto em qq empresa, sem necessidade de fazer uma graduação inteira no assunto.

Informações práticas:

  1. A matéria é a optativa com sigla PTC 2666 até 2016. Não sei a sigla em 2017. O único pré-requisito é Controle I ou equivalente. Acho razoável manter esse pré-requisito para facilitar a interação entre os colegas.
  2. A aprovação continua sendo pelo critério média de 3 notas: auto-avaliação, avaliação dos colegas, e avaliação do professor. Em geral as 3 são idênticas, facilitando as contas.
  3. Minha ideia é escolher um MOOC e fazermos todos juntos ao longo do semestre. Nas aulas discutimos o assunto visto online, e outros tópicos de interesse. Tarefas para antes do início das aulas incluem escolher o curso, a linguagem de programação a ser usada, e um livro de referência se houver.
  4. Por favor me informem o interesse para facilitar o planejamento. Podemos desde já ir colecionando sugestões.
  5. "Data science", "big data", "machine learning" não são os termos mais elegantes da língua de Camões. Por enquanto ficam como guarda-lugares provisórios, a serem substituídos quando chegar um consenso.
Aguardo retorno dos alunos interessados!

04 maio 2017

How Much Does the U.S. Government Subsidize Electricity Generating Technologies?

In reply to

How Much Does the U.S. Government Subsidize Electricity Generating Technologies?

I wrote:

I am sorry, your methodology is just plain wrong.

The majority of US military spending serves to defend easy and inexpensive access to fossil fuel supplies, or to defend the country and our allies against attacks funded by oil revenues. This subsidy is orders of magnitude larger than the ones you did account for.

The subsidy implicit in the environmental and health costs of fossil fuels is even larger. You haven't accounted for those either. Your comparison is not in sense useful as a policy guide.

02 maio 2017

Transfer function zeros

I'm not so sure we really understand them. One the one hand, zeros are more fundamental than poles - poles can be reassigned by feedback, while zeros cannot be moved, they stay right where they are. Yes, stable zeros can be canceled - but they don't really go away, only become hard to observe - or unobservable if you believe in fairies and exact pole-zero cancelations.

On the other hand the very existence of a zero is somewhat arbitrary. A pure delay has no transmission zeros - what comes in gets out, at any input frequency. But a finite-dimensional approximation to a delay has transmission zeros. How can something apparently so immovable appear only in approximations?

The question is of course about the definition of the zeros in terms of the most fundamental description of a linear system, as a convolution of a kernel of integration with an input signal. Definitions in terms of transfer functions or state-space realizations are available and coincide.

Perhaps the definition of a zero as a transmission blocker is not so fundamental. I don't know, and I'm not terribly confident anyone does. Am I wrong?

30 janeiro 2017

Constitutional crisis, coup d'état, and Nathan Hale

It seems that the regime has declared its intention to disregard court orders. Although the Department Of Homeland Security Response To Recent Litigation issued today pays lip service to court orders, the mention is bracketed inside several paragraphs suggesting the opposite. If they do so - I have no manner of knowing what is in the regime's mind - we have a constitutional crisis in the 2nd week of the administration. Sooner than expected, and I come from the background of knowing that democracies can collapse. I expect that the courts will hold for a while, less faith in the Congress.

I am lowering my estimate of how much time we have before an attempted coup d'état to 7 months. How it will play out I have no manner of guessing, but it all makes one think of Nathan Hale.

26 janeiro 2017

How to resist tyranny

One thing is how we educate against bigotry and hatred and ignorance. It is a long term fight. Best left to schools while we can count on them.

The other is how to prevent the country from sliding into an autocracy, which is very immediate - the Party will make all efforts to suppress votes and fraud elections. I am not sure we have even 2 years before the damage is irreversible. Not enough time to educate people.

The only path is to show the voters that they are being taken for idiots. That the plutocrats intend to poison their children's water with lead the way they did to "those people" - forget about global warming, the point of gagging the EPA is so that you don't know that you are drinking the same water they gave to the kids in Flint. That once the billionaires are done with putting for-profit charter schools in inner cities, they will come after your suburban and rural schools as well. That once they are done suppressing the votes of minorities they'll come after the next group. And that once you lost your right to vote your guns won't protect you anymore.

Which is all true, and could perhaps be put in a more intellectual fashion. But voters need to be convinced, and that's the language they'll understand. Is this living with bigotry? Yes but it's either that or tyranny. We don't have enough time to wait until people are kinder, gentler, and more rational.

22 janeiro 2017

The apparatus of revenge and repression will accelerate methodically

Quoting • David Frum ‏@davidfrum  9 Nov 2016
1) Unlike his analogues in Poland/Hungary/Turkey, I don’t imagine that Donald Trump will immediately set out to build an authoritarian state
2) I expect his first priority will be to use the presidency massively to enrich himself.
3) That program of massive self-enrichment however will trigger media investigations and criticism by congressional Democrats.
4) As we’ve seen, Trump cannot tolerate criticism. He prides himself on always retaliating against perceived enemies, by means fair or foul
5) We’ve seen too that Trump’s advisers and aides share this belief. Chris Christie; Corey Lewandowski - they live by gangster morality.
6) So the abuses will start as payback. With a compliant Gop majority in Congress, Trump admin can rewrite laws to enable payback
7) The courts may be an obstacle. But w a compliant Senate, a president can change the courts - as happened in Poland & Hungary.
8) Current IRS commissioner’s term runs to end 2018. But (I looked it up) few commissioners serve the full 5 years.
9) The FBI seems already to have been pre-politicized in Trump’s favor. If Comey resigns, Trump will have another opportunity.
10) Construction of the apparatus of revenge and repression will begin opportunistically & haphazardly. It will accelerate methodically. END

Posted here for future reference. You are warned.

09 janeiro 2017

Now we know how it feels

Those of us who love a democracy, a republic, liberty and freedom in a country under the rule of law, with legal equality and the same rights for everybody, fear we will lose our country - that it will become an authoritarian, lawless, corrupt state, run by thieves, racists, and foreign agents who violently enforce arbitrary decisions for their own personal benefit.

At least we know how the other side felt 8 years ago about losing what they thought their country was.

06 janeiro 2017

Vitriol and unrest

If the United States was divided in sectarian lines - say the middle and South of the country were populated by followers of fundamentalist sects, while the costal regions were mostly inhabited by adepts of mainstream Christianity, with significant Jewish, nonbeliever, and other minorities - then a Syrian-style civil war would be a distinct possibility.

All that would be needed is a tremendous amount of hatred such as that seen in right-wing blogs, and an unstable, corrupt, and authoritarian leader.

04 janeiro 2017

Alguns serviços problemáticos na USP

Alguns serviços de informática na #USP estão funcionando de forma pouco conveniente para o pesquisador, e poderiam ser melhorados.

1 - O acesso à VPN é feito através de um programa da Cisco, chamado AnyConnect. Todo computador tem acesso ao serviço de rede virtual integrado ao sistema operacional, de forma que a obrigatoriedade do uso do software proprietário introduz um custo e uma vulnerabilidade extra, sem benefício para o usuário. Uma desvantagem séria é que o AnyConnect exige que a senha seja digitada manualmente a cada conexão. Isso impede o uso de sistemas seguros de administração de senhas e induz o usuário a utilizar senhas curtas, fáceis de digitar e memorizar, de baixa segurança, tanto para acesso à VPN como aos sistemas administrativos da USP.

2 - Os sistemas de informação da USP funcionam muito mal com o Safari mais recente para MacOS mais recente. Uma consulta ao CCE, renomeado STI por obra e graça do Rodas, informa que os sistemas foram testados em alguma versão do MacOS 10.6 de 2009.

3 - Um dos usos mais importantes da VPN é o acesso aos periódicos usando o portal da Capes. Depois que a Capes forçou todos os acessos a passarem por seu portal, com o objetivo de controlar e centralizar a leitura de artigos científicos, as assinaturas de revistas se tornaram difíceis de encontrar. O sistema de buscas é ruim, porque não fica integrado com todas as demais ferramentas de buscas, e as informações a respeito de assinaturas são pouco claras, quando existem. Me parece que às vezes as assinaturas privilegiam revistas caras, pouco lidas, de editoras comerciais, em detrimento de periódicos baratos e de grande interesse, por exemplo o "Chronicle of Higher Education", que aparentemente não assinamos, embora não consiga afirmar com certeza.