Rick Valenzano

Assistant Professor at Toronto Metropolitan University. General admirer of sport, music, film, and whimsy.

[Rick Valenzano @RickValenzano]

We cover topics such as automated planning, search methods, experimental design, etc. Could be interesting to a variety of @CompSciGradTMU students, as well as @FEAStorontometu students, especially those doing robotics, optimization, and operations research

[Rick Valenzano @RickValenzano]

Excited to be teaching my grad course on heuristic search at @TorontoMetSci this fall! See https://t.co/hB11Yuj9IT for more details.

Note, the course was rescheduled from the winter to the fall. @CompSciGradTMU @TMUgraduate @FEAStorontometu @researchTMU

[Alona Fyshe (she/her) @alonamarie]

I was evacuated from #Jasper while camping with my husband and two young kids this week. This is my #jasperwildfire story. We arrived at Wabasso campground Sunday afternoon. It was 34 degrees, and the man at the check in booth was sweating profusely in the heat. He (1/22)

[Rick Valenzano @RickValenzano]

@c_a_i_a_c Congratulations @SheilaMcIlraith

[Canadian AI Assoc. / Assoc. canadienne pour l'IA @c_a_i_a_c]

This year we are happy to announce two recipients of the CAIAC Lifetime Achievement Award:
Dr. Sheila McIlraith (U. Toronto) and Dr. Ghorbani Ali (U. New Brunswick). These awards recognize outstanding research excellence in AI throughout their careers.

https://t.co/e9PTpZetht

[Rick Valenzano @RickValenzano]

Excited to again be teaching my grad course on Heuristic Search at @TMU_CS . If you are interested, you can find more info at https://t.co/ZFG9IAEVnY

@TMUgraduate @TorontoMetSci @CompSciGradTMU Could also be interesting for Operations Research students in @FEAStorontometu

[Rick Valenzano @RickValenzano]

I am looking for MSc and PhD students to join my team! If you are interested in heuristic search, planning, and the intersection of these topics in RL, consider applying at https://t.co/cpcWcy9E9a . More information about me at https://t.co/AzP9ngvtHd

[Sasho Nikolov ([email protected]) @thesasho]

We have postdoc positions available at U of T’s theory group. I, in particular, would love to have a differential privacy postdoc. Lots of amazing people working in TCS, ML, privacy in the area! Here is the ad https://t.co/UikQbbGQ9v. Feel free to apply or share

[Rick Valenzano @RickValenzano]

But the "heuristic used by Dijkstra's" satisfies that condition.

The bi-directional paper is here: https://t.co/9x0ZkgMzkn

The 1985 paper is here: https://t.co/DhVuMQijxj

[Rick Valenzano @RickValenzano]

There are some differences in setting. Dijkstra finds a path from an input s to ALL other vertices, whereas A* focuses on finding one path from s to a single vertex (typically). This is why the A* work focuses on "must expand" vertices.

[Rick Valenzano @RickValenzano]

@thesasho This was first shown by Dechter and Pearl in 1985. There is more recent work by @nathansttt et al. that I think cleans it up a bit and extends the result to bi-directional search.

[Rick Valenzano @RickValenzano]

And says nothing about those with an evaluation equal to the length of the optimal path. The A*-based work also typically assumes a "consistent" heuristic, which is a triangle inequality on how the heuristic function changes from vertex to vertex

[Rick Valenzano @RickValenzano]

@thesasho This was first shown by Dechter and Pearl in 1985. There is more recent work by @nathansttt et al. that I think cleans it up a bit and extends the result to bi-directional search.

[Rick Valenzano @RickValenzano]

Interesting. I am going to have to take a look. I'm curious how it extends to A*, for which we know that any "equally informed" algorithm B (ie. with the same heuristic), A* will be at least as fast as B (aside from tie-breaking). Dijkstra's after all is just A* with no heuristic

[Sasho Nikolov ([email protected]) @thesasho]

This looks really cool! With the right priority queue, Dijkstra's algorithm is universally optimal: for any graph G and any other algorithm A, there exist weights on the graph edges for which A would make at least as many comparisons as Dijkstra's.

[Rick Valenzano @RickValenzano]

@levilelis @IJCAIconf @JAIR_Editor @RToroIcarte @SheilaMcIlraith Good day for both of us, I see. Congratulations to you too!

[Rick Valenzano @RickValenzano]

Thrilled to announce that our JAIR paper on Reward Machines (https://t.co/xBLyu9mlRD) has won the 2023 IJCAI-JAIR Best Paper Prize @IJCAIconf @JAIR_Editor

Joint work with @RToroIcarte Toryn Klassen and @SheilaMcIlraith

[Rick Valenzano @RickValenzano]

Excited to announce our paper on Learning Reward Machines from experience for partially observable reinforcement learning has been published as a pre-proof in AIJ. Work with @RToroIcarte @mcilraith and other : https://t.co/9H4jqfVGQt

[Rick Valenzano @RickValenzano]

We are still accepting grad applications at @TMU_CS for Canadian students. I myself have a few positions open for people interested in heuristic search, planning, and combining those topics with RL. Check out my work at https://t.co/GvKC6TmNkL if you are interested!

[Rick Valenzano @RickValenzano]

Looking forward to again be teaching a grad class on Heuristic Search and AI at @TorontoMet . If you are a student interested in attending, you can find out more at https://t.co/cYIpmzDQwG @CompSciGradTMU

[Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) @rao2z]

The impressive deep pattern recognition abilities of #DNN's such as #LLM's are sometimes confused for reasoning abilities

I can learn to guess, with high accuracy, whether a SAT instance is satisfiable or not, but this not the same as knowing how to solve SAT. Let me explain. 1/