ECCC-Report TR20-058https://eccc.weizmann.ac.il/report/2020/058Comments and Revisions published for TR20-058en-usSat, 09 Jan 2021 04:12:48 +0200
Revision 1
| Interactive Proofs for Verifying Machine Learning |
Jonathan Shafer,
Amir Yehudayoff,
Shafi Goldwasser,
Guy Rothblum
https://eccc.weizmann.ac.il/report/2020/058#revision1We consider the following question: using a source of labeled data and interaction with an untrusted prover, what is the complexity of verifying that a given hypothesis is "approximately correct"? We study interactive proof systems for PAC verification, where a verifier that interacts with a prover is required to accept good hypotheses, and reject bad hypotheses. Both the verifier and the prover are efficient and have access to data samples from an unknown distribution. We are interested in cases where the verifier can use significantly less data than is required for (agnostic) PAC learning, or use a substantially cheaper data source (e.g., using only random samples for verification, even though learning requires membership queries). We believe that today, when data and data-driven algorithms are quickly gaining prominence, the question of verifying purported outcomes of data analyses is very well-motivated.
We show three main results. First, we prove that for a specific hypothesis class, verification is significantly cheaper than learning in terms of the number of random samples required, even if the verifier engages with the prover only in a single-round (NP-like) protocol. Moreover, for this class we prove that single-round verification is also significantly cheaper than testing closeness to the class. Second, for the broad class of Fourier-sparse boolean functions, we show a multi-round (IP-like) verification protocol, where the prover uses membership queries, and the verifier is able to assess the result while only using random samples. Third, we show that verification is not always more efficient. Namely, we show a class of functions where verification requires as many samples as learning does, up to a logarithmic factor.Sat, 09 Jan 2021 04:12:48 +0200https://eccc.weizmann.ac.il/report/2020/058#revision1
Paper TR20-058
| Interactive Proofs for Verifying Machine Learning |
Jonathan Shafer,
Amir Yehudayoff,
Shafi Goldwasser,
Guy Rothblum
https://eccc.weizmann.ac.il/report/2020/058We consider the following question: using a source of labeled data and interaction with an untrusted prover, what is the complexity of verifying that a given hypothesis is "approximately correct"? We study interactive proof systems for PAC verification, where a verifier that interacts with a prover is required to accept good hypotheses, and reject bad hypotheses. Both the verifier and the prover are efficient and have access to data samples from an unknown distribution. We are interested in cases where the verifier can use significantly less data than is required for (agnostic) PAC learning, or use a substantially cheaper data source (e.g., using only random samples for verification, even though learning requires membership queries). We believe that today, when data and data-driven algorithms are quickly gaining prominence, the question of verifying purported outcomes of data analyses is very well-motivated.
We show three main results. First, we prove that for a specific hypothesis class, verification is significantly cheaper than learning in terms of the number of random samples required, even if the verifier engages with the prover only in a single-round (NP-like) protocol. Moreover, for this class we prove that single-round verification is also significantly cheaper than testing closeness to the class. Second, for the broad class of Fourier-sparse boolean functions, we show a multi-round (IP-like) verification protocol, where the prover uses membership queries, and the verifier is able to assess the result while only using random samples. Third, we show that verification is not always more efficient. Namely, we show a class of functions where verification requires as many samples as learning does, up to a logarithmic factor.Mon, 27 Apr 2020 04:42:43 +0300https://eccc.weizmann.ac.il/report/2020/058